You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Eric Gillespie <ep...@google.com> on 2007/11/13 00:57:03 UTC
WC corruption after checkout on Mac
I have:
0 trunk-r27779% uname -a
Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
0 trunk-r27779% mount
/dev/disk0s2 on / (local, journaled)
and build with:
./configure --prefix=$prefix --disable-static --disable-mod-activation --disable-subdir-config --with-apr=$prefix --with-apr-util=$prefix --with-apxs=$prefix/bin/apxs --with-neon=$prefix --disable-neon-version-check
make -j 4 local-all
sudo make local-install
svn checkout of Subversion, either with ra-neon or ra-svn,
results in random files being 0-bytes. The entries file thinks
it should be 0 bytes, but the text-base has the correct contents.
svn status thinks nothing is wrong until you 'touch' the file;
then it sees that all lines have been removed.
0 trunk-r27779% svn co http://svn.collab.net/repos/svn/trunk -q
0 trunk-r27779% find trunk -size 0
trunk/build/buildcheck.sh
trunk/build/generator/.svn/text-base/__init__.py.svn-base
trunk/build/generator/__init__.py
trunk/build/generator/util/.svn/text-base/__init__.py.svn-base
trunk/build/generator/util/__init__.py
trunk/subversion/bindings/javahl/doc/.svn/text-base/index.html.svn-base
trunk/subversion/bindings/javahl/doc/index.html
trunk/subversion/bindings/swig/python/tests/trac/.svn/text-base/__init__.py.svn-base
trunk/subversion/bindings/swig/python/tests/trac/__init__.py
trunk/subversion/libsvn_client/relocate.c
trunk/subversion/svn/cl.h
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn--help_stderr.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn--version--quiet_stderr.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn--version_stderr.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help--version_stderr.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help_bogus-cmd_stdout.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help_log_switch_stderr.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help_stderr.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_stdout.svn-base
trunk/subversion/tests/cmdline/getopt_tests_data/svn--help_stderr
trunk/subversion/tests/cmdline/getopt_tests_data/svn--version--quiet_stderr
trunk/subversion/tests/cmdline/getopt_tests_data/svn--version_stderr
trunk/subversion/tests/cmdline/getopt_tests_data/svn_help--version_stderr
trunk/subversion/tests/cmdline/getopt_tests_data/svn_help_bogus-cmd_stdout
trunk/subversion/tests/cmdline/getopt_tests_data/svn_help_log_switch_stderr
trunk/subversion/tests/cmdline/getopt_tests_data/svn_help_stderr
trunk/subversion/tests/cmdline/getopt_tests_data/svn_stdout
trunk/www/links.html
0 trunk-r27779% svn st trunk
0 trunk-r27779% touch trunk/build/buildcheck.sh
0 trunk-r27779% svn st trunk
M trunk/build/buildcheck.sh
Oddly enough, glasser cannot reproduce this on his own Mac, which
has at least the same kernel version.
Any ideas?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
Perhaps we should revert r17238? And look closely at all uses of
svn_io_file_del_on_pool_cleanup...
--dave
On Nov 12, 2007 8:52 PM, David Glasser <gl...@davidglasser.net> wrote:
>
> On Nov 12, 2007 7:23 PM, David Glasser <gl...@davidglasser.net> wrote:
> >
> > On Nov 12, 2007 7:22 PM, David Glasser <gl...@davidglasser.net> wrote:
> > > On Nov 12, 2007 5:11 PM, Eric Gillespie <ep...@google.com> wrote:
> > > > Eric Gillespie <ep...@google.com> writes:
> > > >
> > > > > I have:
> > > > >
> > > > > 0 trunk-r27779% uname -a
> > > > > Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
> > > > > 0 trunk-r27779% mount
> > > > > /dev/disk0s2 on / (local, journaled)
> > > >
> > > > Also of interest:
> > > >
> > > > 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> > > > 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> > > > FAIL: lt-translate-test 18: expand author and rev
> > > > FAIL: lt-translate-test 33: unexpand rev and url
> > > > FAIL: lt-translate-test 35: lf_to_crlf; unexpand author
> > >
> > > So, what Eric is reporting is seemingly-random test failures in translate-test.
> > >
> > > I have binary-searched this issue (on the same machine Eric found this
> > > on) and it appears that before r17283, they consistently passed; after
> > > they consistently fail.
> >
> > (By "consistently fail" I mean "consistently, a random subset fails".)
> >
> >
> > > Unfortunately, r17283 is just a test change.
> >
> > --dave
> >
> >
> > --
> > David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
> >
>
> This fixes it. Dunno why.
>
> Index: subversion/libsvn_subr/subst.c
> ===================================================================
> --- subversion/libsvn_subr/subst.c (revision 27787)
> +++ subversion/libsvn_subr/subst.c (working copy)
> @@ -1727,7 +1727,7 @@
> /* For atomicity, we translate to a tmp file and
> then rename the tmp file over the real destination. */
> SVN_ERR(svn_io_open_unique_file2(&d, &dst_tmp, dst,
> - ".tmp", svn_io_file_del_on_pool_cleanup,
> + ".tmp", svn_io_file_del_none,
> pool));
>
> /* Now convert our two open files into streams. */
>
>
>
> --
> David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
>
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
On Nov 12, 2007 7:23 PM, David Glasser <gl...@davidglasser.net> wrote:
>
> On Nov 12, 2007 7:22 PM, David Glasser <gl...@davidglasser.net> wrote:
> > On Nov 12, 2007 5:11 PM, Eric Gillespie <ep...@google.com> wrote:
> > > Eric Gillespie <ep...@google.com> writes:
> > >
> > > > I have:
> > > >
> > > > 0 trunk-r27779% uname -a
> > > > Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
> > > > 0 trunk-r27779% mount
> > > > /dev/disk0s2 on / (local, journaled)
> > >
> > > Also of interest:
> > >
> > > 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> > > 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> > > FAIL: lt-translate-test 18: expand author and rev
> > > FAIL: lt-translate-test 33: unexpand rev and url
> > > FAIL: lt-translate-test 35: lf_to_crlf; unexpand author
> >
> > So, what Eric is reporting is seemingly-random test failures in translate-test.
> >
> > I have binary-searched this issue (on the same machine Eric found this
> > on) and it appears that before r17283, they consistently passed; after
> > they consistently fail.
>
> (By "consistently fail" I mean "consistently, a random subset fails".)
>
>
> > Unfortunately, r17283 is just a test change.
>
> --dave
>
>
> --
> David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
>
This fixes it. Dunno why.
Index: subversion/libsvn_subr/subst.c
===================================================================
--- subversion/libsvn_subr/subst.c (revision 27787)
+++ subversion/libsvn_subr/subst.c (working copy)
@@ -1727,7 +1727,7 @@
/* For atomicity, we translate to a tmp file and
then rename the tmp file over the real destination. */
SVN_ERR(svn_io_open_unique_file2(&d, &dst_tmp, dst,
- ".tmp", svn_io_file_del_on_pool_cleanup,
+ ".tmp", svn_io_file_del_none,
pool));
/* Now convert our two open files into streams. */
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
On Nov 12, 2007 7:22 PM, David Glasser <gl...@davidglasser.net> wrote:
> On Nov 12, 2007 5:11 PM, Eric Gillespie <ep...@google.com> wrote:
> > Eric Gillespie <ep...@google.com> writes:
> >
> > > I have:
> > >
> > > 0 trunk-r27779% uname -a
> > > Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
> > > 0 trunk-r27779% mount
> > > /dev/disk0s2 on / (local, journaled)
> >
> > Also of interest:
> >
> > 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> > 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> > FAIL: lt-translate-test 18: expand author and rev
> > FAIL: lt-translate-test 33: unexpand rev and url
> > FAIL: lt-translate-test 35: lf_to_crlf; unexpand author
>
> So, what Eric is reporting is seemingly-random test failures in translate-test.
>
> I have binary-searched this issue (on the same machine Eric found this
> on) and it appears that before r17283, they consistently passed; after
> they consistently fail.
(By "consistently fail" I mean "consistently, a random subset fails".)
> Unfortunately, r17283 is just a test change.
--dave
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
On Nov 12, 2007 5:11 PM, Eric Gillespie <ep...@google.com> wrote:
> Eric Gillespie <ep...@google.com> writes:
>
> > I have:
> >
> > 0 trunk-r27779% uname -a
> > Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
> > 0 trunk-r27779% mount
> > /dev/disk0s2 on / (local, journaled)
>
> Also of interest:
>
> 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> 1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
> FAIL: lt-translate-test 18: expand author and rev
> FAIL: lt-translate-test 33: unexpand rev and url
> FAIL: lt-translate-test 35: lf_to_crlf; unexpand author
So, what Eric is reporting is seemingly-random test failures in translate-test.
I have binary-searched this issue (on the same machine Eric found this
on) and it appears that before r17283, they consistently passed; after
they consistently fail.
Unfortunately, r17283 is just a test change.
--dave
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Eric Gillespie <ep...@google.com>.
Eric Gillespie <ep...@google.com> writes:
> I have:
>
> 0 trunk-r27779% uname -a
> Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
> 0 trunk-r27779% mount
> /dev/disk0s2 on / (local, journaled)
Also of interest:
1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
1 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
FAIL: lt-translate-test 18: expand author and rev
FAIL: lt-translate-test 33: unexpand rev and url
FAIL: lt-translate-test 35: lf_to_crlf; unexpand author
0 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
FAIL: lt-translate-test 2: convert CRLF to CRLF
FAIL: lt-translate-test 5: convert mixed line endings to CRLF
FAIL: lt-translate-test 7: convert CRLF to LF
0 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
FAIL: lt-translate-test 5: convert mixed line endings to CRLF
FAIL: lt-translate-test 17: expand author and date
FAIL: lt-translate-test 26: cr_to_crlf; expand rev and url
0 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
FAIL: lt-translate-test 8: convert CR to LF
FAIL: lt-translate-test 21: expand author, date, rev, and url
0 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
FAIL: lt-translate-test 12: convert CR to CR
FAIL: lt-translate-test 17: expand author and date
FAIL: lt-translate-test 29: unexpand date
0 trunk-r27779% ./subversion/tests/libsvn_subr/translate-test | grep FAIL
FAIL: lt-translate-test 2: convert CRLF to CRLF
FAIL: lt-translate-test 5: convert mixed line endings to CRLF
FAIL: lt-translate-test 8: convert CR to LF
FAIL: lt-translate-test 16: expand date
FAIL: lt-translate-test 20: expand rev and url
FAIL: lt-translate-test 35: lf_to_crlf; unexpand author
FAIL: lt-translate-test 38: cr_to_crlf; unexpand rev
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
On Nov 13, 2007 4:12 PM, Eric Gillespie <ep...@google.com> wrote:
> - Stop unlinking files we know don't exist
> We know whether the rename failed or not; only unlink if so.
> We should probably audit other del_on_pool_cleanup use, and,
> for that matter, all unlink calls.
In this particular case it's not just "if the rename failed", it's
also if anything fails before the rename (say, the source has
inconsistent newlines). Probably just revert r17238 (or at least the
error usage; dunno about the pool changes).
--dave
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Eric Gillespie <ep...@pretzelnet.org>.
"David Glasser" <gl...@davidglasser.net> writes:
> (One possibility: it is an OS X race condition that is almost
> impossible to trigger; Norton slows things down enough so that it is
> much more likely.)
This is exactly what I think, and the reason I went ahead and
filed the bug with Apple even after you showed turning off Norton
making the bug go away. Norton may be able to cause this to
happen, but ultimately it's the operating system's fault.
Augie and Ben: Are you running Google Desktop, Spotlight, or any
other such program? We have found Google Desktop to be as much a
problem on Windows (where you can't even delete a file someone
like Google Desktop or your antivirus scanner has open).
Augie Fackler <du...@gmail.com> writes:
> Yeah, that's exactly what I see.
>
> On Nov 14, 2007, at 11:57 AM, David Glasser wrote:
>
> > Just to confirm --- the bug is exactly the same, files in a wc end up
> > empty? (And they don't show up in svn status if you don't touch
> > them?)
> >
> > --dave
> >
> > On Nov 14, 2007 7:26 AM, Augie Fackler <du...@gmail.com> wrote:
> >> I've been running it for 25 minutes (flooring both cores) and I'm not
> >> seeing any output. For what it's worth, I'm on 10.5.0, Core2Duo.
Maybe try adding a bunch more load to the mix? Set off a few
make -j 4 builds, a few (sudo find / | sudo xargs cat) > /dev/null 2>&1
and see if you still get nothing.
> >> Do we think there's anything else I can try to nudge it into exposing
> >> the bug? I *know* I've had this problem, I have had to redo a
> >> checkout
> >> about once every 6 months (although, to be fair, that was 10.4.* or
Twice a year is not very often, so it may also be that you'll
have to let this run a while, and continue your other normal work
load, before finally triggering it. But the artificial load
might get you there more quickly.
--
Eric Gillespie <*> epg@pretzelnet.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Augie Fackler <du...@gmail.com>.
Yeah, that's exactly what I see.
On Nov 14, 2007, at 11:57 AM, David Glasser wrote:
> Just to confirm --- the bug is exactly the same, files in a wc end up
> empty? (And they don't show up in svn status if you don't touch
> them?)
>
> --dave
>
> On Nov 14, 2007 7:26 AM, Augie Fackler <du...@gmail.com> wrote:
>> I've been running it for 25 minutes (flooring both cores) and I'm not
>> seeing any output. For what it's worth, I'm on 10.5.0, Core2Duo.
>> Do we think there's anything else I can try to nudge it into exposing
>> the bug? I *know* I've had this problem, I have had to redo a
>> checkout
>> about once every 6 months (although, to be fair, that was 10.4.* or
>> prerelease 10.5 builds, so maybe the bug is fixed now).
>> Peace,
>> Augie
>>
>>
>> On Nov 14, 2007, at 9:35 AM, David Glasser wrote:
>>
>>> Ben, Augie: can you try running Eric's program?
>>>
>>> --dave
>>>
>>> On 11/14/07, Augie Fackler <du...@gmail.com> wrote:
>>>> And I've seen this bug once in a while and I don't even own
>>>> Norton...
>>>>
>>>> On Nov 13, 2007, at 9:00 PM, Ben Collins-Sussman wrote:
>>>>
>>>>> Uh, I've had my Norton Autoprotect switched off since day 1.....
>>>>>
>>>>>
>>>>> On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
>>>>>>
>>>>>> Eric Gillespie <ep...@google.com> writes:
>>>>>>
>>>>>>> Anyone with a Mac, please try this test program.
>>>>>>>
>>>>>>> - save it as bug.c
>>>>>>> - make bug
>>>>>>> - cp -r $svn_trunk/build data
>>>>>>> - while :; do rm -rf work; ./bug; done
>>>>>>>
>>>>>>> Let us know what happens. On one particular Mac we have, we get
>>>>>>> the random 0 byte files:
>>>>>>>
>>>>>>> work/gen_base.py 0 != 39113, unlink_error=2
>>>>>>> work/get-py-info.py 0 != 4255, unlink_error=2
>>>>>>> work/vc6-build.bat.in 0 != 5630, unlink_error=2
>>>>>>> work/gen_make.py 0 != 20457, unlink_error=2
>>>>>>> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
>>>>>>> work/swig.m4 0 != 9023, unlink_error=2
>>>>>>> work/serf.dsp.ezt 0 != 2912, unlink_error=2
>>>>>>> ...
>>>>>>
>>>>>> Glasser is the man, as usual! He figured out that it is
>>>>>> triggered by having "Norton AutoProtect" running. I've filed a
>>>>>> bug with Apple, including everything necessary to reproduce.
>>>>>>
>>>>>> Now, what should *we* do about it? Maybe:
>>>>>>
>>>>>> - Add a FAQ entry
>>>>>> - Stop unlinking files we know don't exist
>>>>>> We know whether the rename failed or not; only unlink if so.
>>>>>> We should probably audit other del_on_pool_cleanup use, and,
>>>>>> for that matter, all unlink calls.
>>>>>>
>>>>>>
>>>>>> ---------------------------------------------------------------------
>>>>>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>>>>>> For additional commands, e-mail: dev-help@subversion.tigris.org
>>>>>>
>>>>>>
>>>>>
>>>>> ---------------------------------------------------------------------
>>>>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>>>>> For additional commands, e-mail: dev-help@subversion.tigris.org
>>>>>
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>>>> For additional commands, e-mail: dev-help@subversion.tigris.org
>>>>
>>>>
>>>
>>>
>>> --
>>> David Glasser | glasser@davidglasser.net | http://
>>> www.davidglasser.net/
>>
>>
>
>
>
> --
> David Glasser | glasser@davidglasser.net | http://
> www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
Just to confirm --- the bug is exactly the same, files in a wc end up
empty? (And they don't show up in svn status if you don't touch
them?)
--dave
On Nov 14, 2007 7:26 AM, Augie Fackler <du...@gmail.com> wrote:
> I've been running it for 25 minutes (flooring both cores) and I'm not
> seeing any output. For what it's worth, I'm on 10.5.0, Core2Duo.
> Do we think there's anything else I can try to nudge it into exposing
> the bug? I *know* I've had this problem, I have had to redo a checkout
> about once every 6 months (although, to be fair, that was 10.4.* or
> prerelease 10.5 builds, so maybe the bug is fixed now).
> Peace,
> Augie
>
>
> On Nov 14, 2007, at 9:35 AM, David Glasser wrote:
>
> > Ben, Augie: can you try running Eric's program?
> >
> > --dave
> >
> > On 11/14/07, Augie Fackler <du...@gmail.com> wrote:
> >> And I've seen this bug once in a while and I don't even own Norton...
> >>
> >> On Nov 13, 2007, at 9:00 PM, Ben Collins-Sussman wrote:
> >>
> >>> Uh, I've had my Norton Autoprotect switched off since day 1.....
> >>>
> >>>
> >>> On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
> >>>>
> >>>> Eric Gillespie <ep...@google.com> writes:
> >>>>
> >>>>> Anyone with a Mac, please try this test program.
> >>>>>
> >>>>> - save it as bug.c
> >>>>> - make bug
> >>>>> - cp -r $svn_trunk/build data
> >>>>> - while :; do rm -rf work; ./bug; done
> >>>>>
> >>>>> Let us know what happens. On one particular Mac we have, we get
> >>>>> the random 0 byte files:
> >>>>>
> >>>>> work/gen_base.py 0 != 39113, unlink_error=2
> >>>>> work/get-py-info.py 0 != 4255, unlink_error=2
> >>>>> work/vc6-build.bat.in 0 != 5630, unlink_error=2
> >>>>> work/gen_make.py 0 != 20457, unlink_error=2
> >>>>> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
> >>>>> work/swig.m4 0 != 9023, unlink_error=2
> >>>>> work/serf.dsp.ezt 0 != 2912, unlink_error=2
> >>>>> ...
> >>>>
> >>>> Glasser is the man, as usual! He figured out that it is
> >>>> triggered by having "Norton AutoProtect" running. I've filed a
> >>>> bug with Apple, including everything necessary to reproduce.
> >>>>
> >>>> Now, what should *we* do about it? Maybe:
> >>>>
> >>>> - Add a FAQ entry
> >>>> - Stop unlinking files we know don't exist
> >>>> We know whether the rename failed or not; only unlink if so.
> >>>> We should probably audit other del_on_pool_cleanup use, and,
> >>>> for that matter, all unlink calls.
> >>>>
> >>>>
> >>>> ---------------------------------------------------------------------
> >>>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> >>>> For additional commands, e-mail: dev-help@subversion.tigris.org
> >>>>
> >>>>
> >>>
> >>> ---------------------------------------------------------------------
> >>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> >>> For additional commands, e-mail: dev-help@subversion.tigris.org
> >>>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> >> For additional commands, e-mail: dev-help@subversion.tigris.org
> >>
> >>
> >
> >
> > --
> > David Glasser | glasser@davidglasser.net | http://
> > www.davidglasser.net/
>
>
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Augie Fackler <du...@gmail.com>.
I've been running it for 25 minutes (flooring both cores) and I'm not
seeing any output. For what it's worth, I'm on 10.5.0, Core2Duo.
Do we think there's anything else I can try to nudge it into exposing
the bug? I *know* I've had this problem, I have had to redo a checkout
about once every 6 months (although, to be fair, that was 10.4.* or
prerelease 10.5 builds, so maybe the bug is fixed now).
Peace,
Augie
On Nov 14, 2007, at 9:35 AM, David Glasser wrote:
> Ben, Augie: can you try running Eric's program?
>
> --dave
>
> On 11/14/07, Augie Fackler <du...@gmail.com> wrote:
>> And I've seen this bug once in a while and I don't even own Norton...
>>
>> On Nov 13, 2007, at 9:00 PM, Ben Collins-Sussman wrote:
>>
>>> Uh, I've had my Norton Autoprotect switched off since day 1.....
>>>
>>>
>>> On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
>>>>
>>>> Eric Gillespie <ep...@google.com> writes:
>>>>
>>>>> Anyone with a Mac, please try this test program.
>>>>>
>>>>> - save it as bug.c
>>>>> - make bug
>>>>> - cp -r $svn_trunk/build data
>>>>> - while :; do rm -rf work; ./bug; done
>>>>>
>>>>> Let us know what happens. On one particular Mac we have, we get
>>>>> the random 0 byte files:
>>>>>
>>>>> work/gen_base.py 0 != 39113, unlink_error=2
>>>>> work/get-py-info.py 0 != 4255, unlink_error=2
>>>>> work/vc6-build.bat.in 0 != 5630, unlink_error=2
>>>>> work/gen_make.py 0 != 20457, unlink_error=2
>>>>> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
>>>>> work/swig.m4 0 != 9023, unlink_error=2
>>>>> work/serf.dsp.ezt 0 != 2912, unlink_error=2
>>>>> ...
>>>>
>>>> Glasser is the man, as usual! He figured out that it is
>>>> triggered by having "Norton AutoProtect" running. I've filed a
>>>> bug with Apple, including everything necessary to reproduce.
>>>>
>>>> Now, what should *we* do about it? Maybe:
>>>>
>>>> - Add a FAQ entry
>>>> - Stop unlinking files we know don't exist
>>>> We know whether the rename failed or not; only unlink if so.
>>>> We should probably audit other del_on_pool_cleanup use, and,
>>>> for that matter, all unlink calls.
>>>>
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>>>> For additional commands, e-mail: dev-help@subversion.tigris.org
>>>>
>>>>
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>>> For additional commands, e-mail: dev-help@subversion.tigris.org
>>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>> For additional commands, e-mail: dev-help@subversion.tigris.org
>>
>>
>
>
> --
> David Glasser | glasser@davidglasser.net | http://
> www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Ben Collins-Sussman <su...@red-bean.com>.
I will do so, as soon as I get home to my Mac. :-)
On Nov 14, 2007 8:35 AM, David Glasser <gl...@davidglasser.net> wrote:
> Ben, Augie: can you try running Eric's program?
>
> --dave
>
>
> On 11/14/07, Augie Fackler <du...@gmail.com> wrote:
> > And I've seen this bug once in a while and I don't even own Norton...
> >
> > On Nov 13, 2007, at 9:00 PM, Ben Collins-Sussman wrote:
> >
> > > Uh, I've had my Norton Autoprotect switched off since day 1.....
> > >
> > >
> > > On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
> > >>
> > >> Eric Gillespie <ep...@google.com> writes:
> > >>
> > >>> Anyone with a Mac, please try this test program.
> > >>>
> > >>> - save it as bug.c
> > >>> - make bug
> > >>> - cp -r $svn_trunk/build data
> > >>> - while :; do rm -rf work; ./bug; done
> > >>>
> > >>> Let us know what happens. On one particular Mac we have, we get
> > >>> the random 0 byte files:
> > >>>
> > >>> work/gen_base.py 0 != 39113, unlink_error=2
> > >>> work/get-py-info.py 0 != 4255, unlink_error=2
> > >>> work/vc6-build.bat.in 0 != 5630, unlink_error=2
> > >>> work/gen_make.py 0 != 20457, unlink_error=2
> > >>> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
> > >>> work/swig.m4 0 != 9023, unlink_error=2
> > >>> work/serf.dsp.ezt 0 != 2912, unlink_error=2
> > >>> ...
> > >>
> > >> Glasser is the man, as usual! He figured out that it is
> > >> triggered by having "Norton AutoProtect" running. I've filed a
> > >> bug with Apple, including everything necessary to reproduce.
> > >>
> > >> Now, what should *we* do about it? Maybe:
> > >>
> > >> - Add a FAQ entry
> > >> - Stop unlinking files we know don't exist
> > >> We know whether the rename failed or not; only unlink if so.
> > >> We should probably audit other del_on_pool_cleanup use, and,
> > >> for that matter, all unlink calls.
> > >>
> > >>
> > >> ---------------------------------------------------------------------
> > >> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> > >> For additional commands, e-mail: dev-help@subversion.tigris.org
> > >>
> > >>
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> > > For additional commands, e-mail: dev-help@subversion.tigris.org
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> > For additional commands, e-mail: dev-help@subversion.tigris.org
> >
> >
>
>
> --
>
> David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
(One possibility: it is an OS X race condition that is almost
impossible to trigger; Norton slows things down enough so that it is
much more likely.)
--dave
On 11/14/07, David Glasser <gl...@davidglasser.net> wrote:
> Ben, Augie: can you try running Eric's program?
>
> --dave
>
> On 11/14/07, Augie Fackler <du...@gmail.com> wrote:
> > And I've seen this bug once in a while and I don't even own Norton...
> >
> > On Nov 13, 2007, at 9:00 PM, Ben Collins-Sussman wrote:
> >
> > > Uh, I've had my Norton Autoprotect switched off since day 1.....
> > >
> > >
> > > On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
> > >>
> > >> Eric Gillespie <ep...@google.com> writes:
> > >>
> > >>> Anyone with a Mac, please try this test program.
> > >>>
> > >>> - save it as bug.c
> > >>> - make bug
> > >>> - cp -r $svn_trunk/build data
> > >>> - while :; do rm -rf work; ./bug; done
> > >>>
> > >>> Let us know what happens. On one particular Mac we have, we get
> > >>> the random 0 byte files:
> > >>>
> > >>> work/gen_base.py 0 != 39113, unlink_error=2
> > >>> work/get-py-info.py 0 != 4255, unlink_error=2
> > >>> work/vc6-build.bat.in 0 != 5630, unlink_error=2
> > >>> work/gen_make.py 0 != 20457, unlink_error=2
> > >>> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
> > >>> work/swig.m4 0 != 9023, unlink_error=2
> > >>> work/serf.dsp.ezt 0 != 2912, unlink_error=2
> > >>> ...
> > >>
> > >> Glasser is the man, as usual! He figured out that it is
> > >> triggered by having "Norton AutoProtect" running. I've filed a
> > >> bug with Apple, including everything necessary to reproduce.
> > >>
> > >> Now, what should *we* do about it? Maybe:
> > >>
> > >> - Add a FAQ entry
> > >> - Stop unlinking files we know don't exist
> > >> We know whether the rename failed or not; only unlink if so.
> > >> We should probably audit other del_on_pool_cleanup use, and,
> > >> for that matter, all unlink calls.
> > >>
> > >>
> > >> ---------------------------------------------------------------------
> > >> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> > >> For additional commands, e-mail: dev-help@subversion.tigris.org
> > >>
> > >>
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> > > For additional commands, e-mail: dev-help@subversion.tigris.org
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> > For additional commands, e-mail: dev-help@subversion.tigris.org
> >
> >
>
>
> --
> David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
>
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
Ben, Augie: can you try running Eric's program?
--dave
On 11/14/07, Augie Fackler <du...@gmail.com> wrote:
> And I've seen this bug once in a while and I don't even own Norton...
>
> On Nov 13, 2007, at 9:00 PM, Ben Collins-Sussman wrote:
>
> > Uh, I've had my Norton Autoprotect switched off since day 1.....
> >
> >
> > On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
> >>
> >> Eric Gillespie <ep...@google.com> writes:
> >>
> >>> Anyone with a Mac, please try this test program.
> >>>
> >>> - save it as bug.c
> >>> - make bug
> >>> - cp -r $svn_trunk/build data
> >>> - while :; do rm -rf work; ./bug; done
> >>>
> >>> Let us know what happens. On one particular Mac we have, we get
> >>> the random 0 byte files:
> >>>
> >>> work/gen_base.py 0 != 39113, unlink_error=2
> >>> work/get-py-info.py 0 != 4255, unlink_error=2
> >>> work/vc6-build.bat.in 0 != 5630, unlink_error=2
> >>> work/gen_make.py 0 != 20457, unlink_error=2
> >>> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
> >>> work/swig.m4 0 != 9023, unlink_error=2
> >>> work/serf.dsp.ezt 0 != 2912, unlink_error=2
> >>> ...
> >>
> >> Glasser is the man, as usual! He figured out that it is
> >> triggered by having "Norton AutoProtect" running. I've filed a
> >> bug with Apple, including everything necessary to reproduce.
> >>
> >> Now, what should *we* do about it? Maybe:
> >>
> >> - Add a FAQ entry
> >> - Stop unlinking files we know don't exist
> >> We know whether the rename failed or not; only unlink if so.
> >> We should probably audit other del_on_pool_cleanup use, and,
> >> for that matter, all unlink calls.
> >>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> >> For additional commands, e-mail: dev-help@subversion.tigris.org
> >>
> >>
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> > For additional commands, e-mail: dev-help@subversion.tigris.org
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
>
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Augie Fackler <du...@gmail.com>.
And I've seen this bug once in a while and I don't even own Norton...
On Nov 13, 2007, at 9:00 PM, Ben Collins-Sussman wrote:
> Uh, I've had my Norton Autoprotect switched off since day 1.....
>
>
> On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
>>
>> Eric Gillespie <ep...@google.com> writes:
>>
>>> Anyone with a Mac, please try this test program.
>>>
>>> - save it as bug.c
>>> - make bug
>>> - cp -r $svn_trunk/build data
>>> - while :; do rm -rf work; ./bug; done
>>>
>>> Let us know what happens. On one particular Mac we have, we get
>>> the random 0 byte files:
>>>
>>> work/gen_base.py 0 != 39113, unlink_error=2
>>> work/get-py-info.py 0 != 4255, unlink_error=2
>>> work/vc6-build.bat.in 0 != 5630, unlink_error=2
>>> work/gen_make.py 0 != 20457, unlink_error=2
>>> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
>>> work/swig.m4 0 != 9023, unlink_error=2
>>> work/serf.dsp.ezt 0 != 2912, unlink_error=2
>>> ...
>>
>> Glasser is the man, as usual! He figured out that it is
>> triggered by having "Norton AutoProtect" running. I've filed a
>> bug with Apple, including everything necessary to reproduce.
>>
>> Now, what should *we* do about it? Maybe:
>>
>> - Add a FAQ entry
>> - Stop unlinking files we know don't exist
>> We know whether the rename failed or not; only unlink if so.
>> We should probably audit other del_on_pool_cleanup use, and,
>> for that matter, all unlink calls.
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>> For additional commands, e-mail: dev-help@subversion.tigris.org
>>
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Ben Collins-Sussman <su...@red-bean.com>.
Uh, I've had my Norton Autoprotect switched off since day 1.....
On Nov 13, 2007 6:12 PM, Eric Gillespie <ep...@google.com> wrote:
>
> Eric Gillespie <ep...@google.com> writes:
>
> > Anyone with a Mac, please try this test program.
> >
> > - save it as bug.c
> > - make bug
> > - cp -r $svn_trunk/build data
> > - while :; do rm -rf work; ./bug; done
> >
> > Let us know what happens. On one particular Mac we have, we get
> > the random 0 byte files:
> >
> > work/gen_base.py 0 != 39113, unlink_error=2
> > work/get-py-info.py 0 != 4255, unlink_error=2
> > work/vc6-build.bat.in 0 != 5630, unlink_error=2
> > work/gen_make.py 0 != 20457, unlink_error=2
> > work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
> > work/swig.m4 0 != 9023, unlink_error=2
> > work/serf.dsp.ezt 0 != 2912, unlink_error=2
> > ...
>
> Glasser is the man, as usual! He figured out that it is
> triggered by having "Norton AutoProtect" running. I've filed a
> bug with Apple, including everything necessary to reproduce.
>
> Now, what should *we* do about it? Maybe:
>
> - Add a FAQ entry
> - Stop unlinking files we know don't exist
> We know whether the rename failed or not; only unlink if so.
> We should probably audit other del_on_pool_cleanup use, and,
> for that matter, all unlink calls.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Eric Gillespie <ep...@google.com>.
Eric Gillespie <ep...@google.com> writes:
> Anyone with a Mac, please try this test program.
>
> - save it as bug.c
> - make bug
> - cp -r $svn_trunk/build data
> - while :; do rm -rf work; ./bug; done
>
> Let us know what happens. On one particular Mac we have, we get
> the random 0 byte files:
>
> work/gen_base.py 0 != 39113, unlink_error=2
> work/get-py-info.py 0 != 4255, unlink_error=2
> work/vc6-build.bat.in 0 != 5630, unlink_error=2
> work/gen_make.py 0 != 20457, unlink_error=2
> work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
> work/swig.m4 0 != 9023, unlink_error=2
> work/serf.dsp.ezt 0 != 2912, unlink_error=2
> ...
Glasser is the man, as usual! He figured out that it is
triggered by having "Norton AutoProtect" running. I've filed a
bug with Apple, including everything necessary to reproduce.
Now, what should *we* do about it? Maybe:
- Add a FAQ entry
- Stop unlinking files we know don't exist
We know whether the rename failed or not; only unlink if so.
We should probably audit other del_on_pool_cleanup use, and,
for that matter, all unlink calls.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Stefan Haller <ha...@ableton.com>.
Eric Gillespie <ep...@google.com> wrote:
> - while :; do rm -rf work; ./bug; done
>
> Let us know what happens.
I let this run for a couple of minutes, and didn't get any output. This
is on 10.5.
I did notice pretty heavy spotlight activity though (to be expected).
Could this be related? On your Mac that shows the problem, does it make
any difference if you disable spotlight?
--
Stefan Haller
Ableton
http://www.ableton.com/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Eric Gillespie <ep...@google.com>.
Anyone with a Mac, please try this test program.
- save it as bug.c
- make bug
- cp -r $svn_trunk/build data
- while :; do rm -rf work; ./bug; done
Let us know what happens. On one particular Mac we have, we get
the random 0 byte files:
work/gen_base.py 0 != 39113, unlink_error=2
work/get-py-info.py 0 != 4255, unlink_error=2
work/vc6-build.bat.in 0 != 5630, unlink_error=2
work/gen_make.py 0 != 20457, unlink_error=2
work/svn_config.vcproj.ezt 0 != 2515, unlink_error=2
work/swig.m4 0 != 9023, unlink_error=2
work/serf.dsp.ezt 0 != 2912, unlink_error=2
...
Re: WC corruption after checkout on Mac
Posted by David Glasser <gl...@davidglasser.net>.
On Nov 12, 2007 5:14 PM, Ben Collins-Sussman <su...@red-bean.com> wrote:
> This happened to me yesterday on my mac as well. I ran 'svn up' on my
> svn trunk (which was about a week out of date), and somehow my
> svn_base64.h file became zero bytes in size. 'svn status' reported
> nothing abnormal at all.
Note that the reason that "svn status" is silent is that the
working-size field in the entries file (new in 1.5) is being set based
on the (incorrectly empty) working copy file to 0, and so since it
sees unchanged working-size and timestamp, 'svn st' assumes no change.
--dave
--
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Ben Collins-Sussman <su...@red-bean.com>.
This happened to me yesterday on my mac as well. I ran 'svn up' on my
svn trunk (which was about a week out of date), and somehow my
svn_base64.h file became zero bytes in size. 'svn status' reported
nothing abnormal at all.
I wasn't able to reproduce the bug. But hey, it looks ilke epg has
experienced it too!
On Nov 12, 2007 6:57 PM, Eric Gillespie <ep...@google.com> wrote:
> I have:
>
> 0 trunk-r27779% uname -a
> Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
> 0 trunk-r27779% mount
> /dev/disk0s2 on / (local, journaled)
>
> and build with:
>
> ./configure --prefix=$prefix --disable-static --disable-mod-activation --disable-subdir-config --with-apr=$prefix --with-apr-util=$prefix --with-apxs=$prefix/bin/apxs --with-neon=$prefix --disable-neon-version-check
> make -j 4 local-all
> sudo make local-install
>
> svn checkout of Subversion, either with ra-neon or ra-svn,
> results in random files being 0-bytes. The entries file thinks
> it should be 0 bytes, but the text-base has the correct contents.
> svn status thinks nothing is wrong until you 'touch' the file;
> then it sees that all lines have been removed.
>
> 0 trunk-r27779% svn co http://svn.collab.net/repos/svn/trunk -q
> 0 trunk-r27779% find trunk -size 0
> trunk/build/buildcheck.sh
> trunk/build/generator/.svn/text-base/__init__.py.svn-base
> trunk/build/generator/__init__.py
> trunk/build/generator/util/.svn/text-base/__init__.py.svn-base
> trunk/build/generator/util/__init__.py
> trunk/subversion/bindings/javahl/doc/.svn/text-base/index.html.svn-base
> trunk/subversion/bindings/javahl/doc/index.html
> trunk/subversion/bindings/swig/python/tests/trac/.svn/text-base/__init__.py.svn-base
> trunk/subversion/bindings/swig/python/tests/trac/__init__.py
> trunk/subversion/libsvn_client/relocate.c
> trunk/subversion/svn/cl.h
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn--help_stderr.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn--version--quiet_stderr.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn--version_stderr.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help--version_stderr.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help_bogus-cmd_stdout.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help_log_switch_stderr.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_help_stderr.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/.svn/text-base/svn_stdout.svn-base
> trunk/subversion/tests/cmdline/getopt_tests_data/svn--help_stderr
> trunk/subversion/tests/cmdline/getopt_tests_data/svn--version--quiet_stderr
> trunk/subversion/tests/cmdline/getopt_tests_data/svn--version_stderr
> trunk/subversion/tests/cmdline/getopt_tests_data/svn_help--version_stderr
> trunk/subversion/tests/cmdline/getopt_tests_data/svn_help_bogus-cmd_stdout
> trunk/subversion/tests/cmdline/getopt_tests_data/svn_help_log_switch_stderr
> trunk/subversion/tests/cmdline/getopt_tests_data/svn_help_stderr
> trunk/subversion/tests/cmdline/getopt_tests_data/svn_stdout
> trunk/www/links.html
>
> 0 trunk-r27779% svn st trunk
> 0 trunk-r27779% touch trunk/build/buildcheck.sh
> 0 trunk-r27779% svn st trunk
> M trunk/build/buildcheck.sh
>
> Oddly enough, glasser cannot reproduce this on his own Mac, which
> has at least the same kernel version.
>
> Any ideas?
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Karl Fogel <kf...@red-bean.com>.
Eric Gillespie <ep...@google.com> writes:
> svn checkout of Subversion, either with ra-neon or ra-svn,
> results in random files being 0-bytes. The entries file thinks
> it should be 0 bytes, but the text-base has the correct contents.
> svn status thinks nothing is wrong until you 'touch' the file;
> then it sees that all lines have been removed.
>
> [...]
>
> Oddly enough, glasser cannot reproduce this on his own Mac, which
> has at least the same kernel version.
>
> Any ideas?
Hrm. It sounds like you're the only person with a reliable
reproduction recipe right now. Can you GDB through
subversion/libsvn_wc/update_editor.c:merge_file()
and see where things go wrong?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: WC corruption after checkout on Mac
Posted by Eric Gillespie <ep...@google.com>.
Eric Gillespie <ep...@google.com> writes:
> I have:
>
> 0 trunk-r27779% uname -a
> Darwin 8.10.1 Darwin Kernel Version 8.10.1: Wed May 23 16:33:00 PDT 2007; root:xnu-792.22.5~1/RELEASE_I386 i386 i386
> 0 trunk-r27779% mount
> /dev/disk0s2 on / (local, journaled)
I forgot to note: I can only see this problem on Mac, not Linux
or Windows (all i386).
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org