You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Jon Trowbridge <tr...@ximian.com> on 2001/07/09 02:37:46 UTC
Two problems w/ subversion
OK, now I can build it, but I've found a few glitches
(1) A minor problem.
The test scripts (basic-tests.py in particular) seem to require python
2.0. On my machine (running debian Woody), "python" invokes python
1.5.2 and "python2" runs 2.0. Thus things like basic-tests.py won't run
for me unless I manually run them from the command line with "python2
./basic-tests.py".
(2) A larger problem
svnadmin segfaults when I try to create a repository. Here is a
backtrace
(gdb) run create /home/trow/svn-test-repos
Starting program: /home/trow/bin/svnadmin create
/home/trow/svn-test-repos
Program received signal SIGSEGV, Segmentation fault.
0x0 in ?? ()
(gdb) bt
#0 0x0 in ?? ()
#1 0x804ad77 in cleanup_fs (fs=0x80736a4) at
subversion/libsvn_fs/fs.c:143
#2 0x804b572 in svn_fs_create_berkeley (fs=0x80736a4,
path=0x7ffffa22 "/home/trow/svn-test-repos")
at subversion/libsvn_fs/fs.c:436
#3 0x804a5f3 in main (argc=3, argv=0x7ffff8c4)
at subversion/svnadmin/main.c:159
I'm using the db3 packages provided by debian.
-JT
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Mo DeJong <md...@cygnus.com>.
On 17 Jul 2001, Karl Fogel wrote:
> If that's true, let's look at the same situation with your patch
> applied. You claim ./configure would have given a warning. Why would
> it do that? What would that warning be based on? As far as configure
> can tell, his system does have the correct db, so it shouldn't tell
> him to download the tarball, right?
Oh, I see what you mean. You are correct in that it would not
have caught this case since it would have searched the
system and found a "compatible" db install.
> I'm just confused by your claim that the patch would have helped Jon;
> and worried that perhaps this indicates there's something important
> about your patch that we're not understanding...
I meant that if he had compiled from the source in the db tree,
he would have never run into the problem in the first place.
With the default searching code in place, this would continue
to be a problem. Perhaps that is a good argument for not
using the default system searching code by default, but
that is an argument for another time.
cheers
Mo
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Karl Fogel <kf...@collab.net>.
Mo DeJong <md...@cygnus.com> writes:
> Because, he would have run ./configure, it would have
> printed a warning, he would have pulled down the
> tar ball and then rerun configure. He could also rerun
> configure an pass in a --with-berkeley-db option,
> but the point of my patch is to avoid that. There
> is no modification to ./autogen.sh to require a db
> dir because that would make it hard to use with
> an installed version. That was in fact the reason
> given for rejecting the patch in a earlier form.
Thanks for the explanation.
I'm still confused, though. Sorry if I'm being dense, here's my
confusion:
Jon said (or implied, maybe?) that his Debian system's native db
*already* claimed to be the correct version, that is, the version our
configure script looks for. However, something about the Debian
package was messed up, and so it still didn't work for him when he
actually ran Subversion.
That's my understanding of his post, anyway.
If that's true, let's look at the same situation with your patch
applied. You claim ./configure would have given a warning. Why would
it do that? What would that warning be based on? As far as configure
can tell, his system does have the correct db, so it shouldn't tell
him to download the tarball, right?
None of this is an argument against your patch, by the way. I think
the patch implements a good goal, and we're going to examine it soon,
many apologies for the delay.
I'm just confused by your claim that the patch would have helped Jon;
and worried that perhaps this indicates there's something important
about your patch that we're not understanding...
Best,
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Mo DeJong <md...@cygnus.com>.
On 17 Jul 2001, Karl Fogel wrote:
...
> > The above problem would have been avoided if this patch had
> > been applied. Apr, neon, and expat already build out of the
> > src tree. Why not db?
>
> I don't see how this patch would have solved that problem.
>
> It would solve that problem if the person knew to not use their
> system's pre-installed db3 and instead had their own copy of the
> sources around. But that's exactly what Jon Trowbridge didn't know.
This is the exact situation my patch seeks to avoid. In almost
every case, the version of db installed on the system
is going to be the wrong one. Some folks will want the
option of using the system installed version, but these
will most likely be folks putting together RPMs or DEBS
for "normal users".
> Am I misunderstanding your patch? How would Jon's results have been
> different had your patch been in effect?
Because, he would have run ./configure, it would have
printed a warning, he would have pulled down the
tar ball and then rerun configure. He could also rerun
configure an pass in a --with-berkeley-db option,
but the point of my patch is to avoid that. There
is no modification to ./autogen.sh to require a db
dir because that would make it hard to use with
an installed version. That was in fact the reason
given for rejecting the patch in a earlier form.
cheers
Mo
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Karl Fogel <kf...@collab.net>.
Mo DeJong <md...@cygnus.com> writes:
> On 17 Jul 2001, Jon Trowbridge wrote:
> > An update: I've continued to have this problem, until tonight when I
> > removed the debian libdb3-dev package and built/installed libdb3 from
> > the sleepycat tarball. I've now been able to create a repository, so
> > maybe the debian package is broken.
> >
> > -JT
>
> Could someone take a look at my "build db in the tree" patch?
>
> http://subversion.tigris.org/servlets/ReadMsg?msgId=28881&listName=dev
>
> The above problem would have been avoided if this patch had
> been applied. Apr, neon, and expat already build out of the
> src tree. Why not db?
I don't see how this patch would have solved that problem.
It would solve that problem if the person knew to not use their
system's pre-installed db3 and instead had their own copy of the
sources around. But that's exactly what Jon Trowbridge didn't know.
Am I misunderstanding your patch? How would Jon's results have been
different had your patch been in effect?
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Mo DeJong <md...@cygnus.com>.
On 17 Jul 2001, Jon Trowbridge wrote:
> On 10 Jul 2001 11:15:32 -0500, cmpilato@collab.net wrote:
> > Jon Trowbridge <tr...@ximian.com> writes:
> > > (2) A larger problem
> > >
> > > svnadmin segfaults when I try to create a repository.
> >
> > I know this was happening a couple of weeks ago, but that bug has
> > since been fixed. I use svnadmin daily here to create repositories,
> > and am not seeing the problem. Any chance that your binary is out of
> > date?
>
> An update: I've continued to have this problem, until tonight when I
> removed the debian libdb3-dev package and built/installed libdb3 from
> the sleepycat tarball. I've now been able to create a repository, so
> maybe the debian package is broken.
>
> -JT
Could someone take a look at my "build db in the tree" patch?
http://subversion.tigris.org/servlets/ReadMsg?msgId=28881&listName=dev
The above problem would have been avoided if this patch had
been applied. Apr, neon, and expat already build out of the
src tree. Why not db?
cheers
Mo
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Jon Trowbridge <tr...@ximian.com>.
On 10 Jul 2001 11:15:32 -0500, cmpilato@collab.net wrote:
> Jon Trowbridge <tr...@ximian.com> writes:
> > (2) A larger problem
> >
> > svnadmin segfaults when I try to create a repository.
>
> I know this was happening a couple of weeks ago, but that bug has
> since been fixed. I use svnadmin daily here to create repositories,
> and am not seeing the problem. Any chance that your binary is out of
> date?
An update: I've continued to have this problem, until tonight when I
removed the debian libdb3-dev package and built/installed libdb3 from
the sleepycat tarball. I've now been able to create a repository, so
maybe the debian package is broken.
-JT
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Karl Fogel <kf...@collab.net>.
ender <kt...@earthlink.net> writes:
> ok, i just built my first subversion server, did some testing and realized
> that python2 IS required for the test cases to be portable ( i was testing
> importability before for syntax compatiblity). i'm sincerely sorry for any
> confusion i might have caused. the problem lies in that python1.5.2's popen2
> call is unix only. on windows it doesn't exist.
>
> send the patch to /dev/null
>
> back to lurking
Hey, no problem -- thanks for the attempted patch anyway, and I hope
we'll get more patches from you in the future!
-Karl
> On Wednesday 11 July 2001 06:21, ender wrote:
> >>On Tuesday 10 July 2001 18:04, Karl Fogel wrote:
> >>>>Thanks for the patch, Kapil! Mind if I request you repost with a
> >>>>couple of minor frobs?:
> >>
> >>no problem. new sanitized, unified, and commented patch is attached.
> >>
> >>>> 1. Run diff -u (or diff -c if you prefer), so the patch is more
> >>>> readable.
> >>>>
> >>>> 2. There's no need to leave the old code commented out -- just
> >>>> remove it entirely. We'll have the diff context right there if
> >>>> we want to see the old code, anyway.
> >>>>
> >>>> 3. Write a log message; even a brief one helps a reader get an
> >>>> overview of the patch.
> >>>>
> >>>>
> >>>>Thanks, hope this isn't too much of an imposition on your time...
> >>
> >>not at all, the changes are trivial and the clean up is more in keeping
> >> with the high quality of the subversion source (which seems well
> >> documented and nice overall imo).
> >>
> >>~Kapil
> >>
> >>>>-K
> >>>>
> >>>>kapil thangavelu <kt...@earthlink.net> writes:
> >>>>> On Tuesday 10 July 2001 15:14, Karl Fogel wrote:
> >>>>> >>Thanks for the summary, Trent.
> >>>>> >>
> >>>>> >>Some history here:
> >>>>> >>
> >>>>> >>Subversion does require Python 2.0 for a reason: there's some code in
> >>>>> >>the test suite that uses a feature that wasn't available in 1.5.2. I
> >>>>> >>frankly don't remember exactly what the issue was, I just remember
> >>>>> >>that Mike Sussman or Ben discovered that something didn't work with
> >>>>> >>pre-2.0 Pythons, did a little research, and found out that we'd made
> >>>>> >>use of a feature new in 2.0. In other words, 2.0 turned out to be
> >>>>> >>backward-compatible but not forward-compatible with 1.5.2... as is
> >>>>> >>usually the case with languages, I suppose.
> >>>>>
> >>>>> one more thing that python 2 adds is list comprehensions and some other
> >>>>> syntatic sugar for function calling, which apparently is the problem in
> >>>>> the test suites. it only affects one file actions.py and it looks easy
> >>>>> enough to fix.
> >>>>>
> >>>>> >>SOLUTION:
> >>>>>
> >>>>> i'm attaching the diff of the changes to make it work with python 1.5.2
> >>>>> diff actions.py oldactions.py > actions.diff
> >>>>>
> >>>>> cheers
> >>>>>
> >>>>> kapil
> >>>>
> >>>>---------------------------------------------------------------------
> >>>>To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> >>>>For additional commands, e-mail: dev-help@subversion.tigris.org
>
> ----------------------------------------
> Content-Type: text/plain; charset="iso-8859-1"; name="actions.py.diff"
> Content-Transfer-Encoding: base64
> Content-Description:
> ----------------------------------------
>
> ----------------------------------------
> Content-Type: text/plain; charset="us-ascii"; name="Attachment: 2"
> Content-Transfer-Encoding: 7bit
> Content-Description:
> ----------------------------------------
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by ender <kt...@earthlink.net>.
ok, i just built my first subversion server, did some testing and realized
that python2 IS required for the test cases to be portable ( i was testing
importability before for syntax compatiblity). i'm sincerely sorry for any
confusion i might have caused. the problem lies in that python1.5.2's popen2
call is unix only. on windows it doesn't exist.
send the patch to /dev/null
back to lurking
cheers
kapil
On Wednesday 11 July 2001 06:21, ender wrote:
>>On Tuesday 10 July 2001 18:04, Karl Fogel wrote:
>>>>Thanks for the patch, Kapil! Mind if I request you repost with a
>>>>couple of minor frobs?:
>>
>>no problem. new sanitized, unified, and commented patch is attached.
>>
>>>> 1. Run diff -u (or diff -c if you prefer), so the patch is more
>>>> readable.
>>>>
>>>> 2. There's no need to leave the old code commented out -- just
>>>> remove it entirely. We'll have the diff context right there if
>>>> we want to see the old code, anyway.
>>>>
>>>> 3. Write a log message; even a brief one helps a reader get an
>>>> overview of the patch.
>>>>
>>>>
>>>>Thanks, hope this isn't too much of an imposition on your time...
>>
>>not at all, the changes are trivial and the clean up is more in keeping
>> with the high quality of the subversion source (which seems well
>> documented and nice overall imo).
>>
>>~Kapil
>>
>>>>-K
>>>>
>>>>kapil thangavelu <kt...@earthlink.net> writes:
>>>>> On Tuesday 10 July 2001 15:14, Karl Fogel wrote:
>>>>> >>Thanks for the summary, Trent.
>>>>> >>
>>>>> >>Some history here:
>>>>> >>
>>>>> >>Subversion does require Python 2.0 for a reason: there's some code in
>>>>> >>the test suite that uses a feature that wasn't available in 1.5.2. I
>>>>> >>frankly don't remember exactly what the issue was, I just remember
>>>>> >>that Mike Sussman or Ben discovered that something didn't work with
>>>>> >>pre-2.0 Pythons, did a little research, and found out that we'd made
>>>>> >>use of a feature new in 2.0. In other words, 2.0 turned out to be
>>>>> >>backward-compatible but not forward-compatible with 1.5.2... as is
>>>>> >>usually the case with languages, I suppose.
>>>>>
>>>>> one more thing that python 2 adds is list comprehensions and some other
>>>>> syntatic sugar for function calling, which apparently is the problem in
>>>>> the test suites. it only affects one file actions.py and it looks easy
>>>>> enough to fix.
>>>>>
>>>>> >>SOLUTION:
>>>>>
>>>>> i'm attaching the diff of the changes to make it work with python 1.5.2
>>>>> diff actions.py oldactions.py > actions.diff
>>>>>
>>>>> cheers
>>>>>
>>>>> kapil
>>>>
>>>>---------------------------------------------------------------------
>>>>To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>>>>For additional commands, e-mail: dev-help@subversion.tigris.org
----------------------------------------
Content-Type: text/plain; charset="iso-8859-1"; name="actions.py.diff"
Content-Transfer-Encoding: base64
Content-Description:
----------------------------------------
----------------------------------------
Content-Type: text/plain; charset="us-ascii"; name="Attachment: 2"
Content-Transfer-Encoding: 7bit
Content-Description:
----------------------------------------
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by ender <kt...@earthlink.net>.
On Tuesday 10 July 2001 18:04, Karl Fogel wrote:
>>Thanks for the patch, Kapil! Mind if I request you repost with a
>>couple of minor frobs?:
no problem. new sanitized, unified, and commented patch is attached.
>> 1. Run diff -u (or diff -c if you prefer), so the patch is more
>> readable.
>>
>> 2. There's no need to leave the old code commented out -- just
>> remove it entirely. We'll have the diff context right there if
>> we want to see the old code, anyway.
>>
>> 3. Write a log message; even a brief one helps a reader get an
>> overview of the patch.
>>Thanks, hope this isn't too much of an imposition on your time...
not at all, the changes are trivial and the clean up is more in keeping with
the high quality of the subversion source (which seems well documented and
nice overall imo).
~Kapil
>>-K
>>
>>kapil thangavelu <kt...@earthlink.net> writes:
>>> On Tuesday 10 July 2001 15:14, Karl Fogel wrote:
>>> >>Thanks for the summary, Trent.
>>> >>
>>> >>Some history here:
>>> >>
>>> >>Subversion does require Python 2.0 for a reason: there's some code in
>>> >>the test suite that uses a feature that wasn't available in 1.5.2. I
>>> >>frankly don't remember exactly what the issue was, I just remember
>>> >>that Mike Sussman or Ben discovered that something didn't work with
>>> >>pre-2.0 Pythons, did a little research, and found out that we'd made
>>> >>use of a feature new in 2.0. In other words, 2.0 turned out to be
>>> >>backward-compatible but not forward-compatible with 1.5.2... as is
>>> >>usually the case with languages, I suppose.
>>>
>>> one more thing that python 2 adds is list comprehensions and some other
>>> syntatic sugar for function calling, which apparently is the problem in
>>> the test suites. it only affects one file actions.py and it looks easy
>>> enough to fix.
>>>
>>> >>SOLUTION:
>>>
>>> i'm attaching the diff of the changes to make it work with python 1.5.2
>>> diff actions.py oldactions.py > actions.diff
>>>
>>> cheers
>>>
>>> kapil
>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
>>For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Karl Fogel <kf...@collab.net>.
Thanks for the patch, Kapil! Mind if I request you repost with a
couple of minor frobs?:
1. Run diff -u (or diff -c if you prefer), so the patch is more
readable.
2. There's no need to leave the old code commented out -- just
remove it entirely. We'll have the diff context right there if
we want to see the old code, anyway.
3. Write a log message; even a brief one helps a reader get an
overview of the patch.
Thanks, hope this isn't too much of an imposition on your time...
-K
kapil thangavelu <kt...@earthlink.net> writes:
> On Tuesday 10 July 2001 15:14, Karl Fogel wrote:
> >>Thanks for the summary, Trent.
> >>
> >>Some history here:
> >>
> >>Subversion does require Python 2.0 for a reason: there's some code in
> >>the test suite that uses a feature that wasn't available in 1.5.2. I
> >>frankly don't remember exactly what the issue was, I just remember
> >>that Mike Sussman or Ben discovered that something didn't work with
> >>pre-2.0 Pythons, did a little research, and found out that we'd made
> >>use of a feature new in 2.0. In other words, 2.0 turned out to be
> >>backward-compatible but not forward-compatible with 1.5.2... as is
> >>usually the case with languages, I suppose.
>
> one more thing that python 2 adds is list comprehensions and some other
> syntatic sugar for function calling, which apparently is the problem in the
> test suites. it only affects one file actions.py and it looks easy enough to
> fix.
>
> >>SOLUTION:
>
> i'm attaching the diff of the changes to make it work with python 1.5.2
> diff actions.py oldactions.py > actions.diff
>
> cheers
>
> kapil
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by kapil thangavelu <kt...@earthlink.net>.
On Tuesday 10 July 2001 15:14, Karl Fogel wrote:
>>Thanks for the summary, Trent.
>>
>>Some history here:
>>
>>Subversion does require Python 2.0 for a reason: there's some code in
>>the test suite that uses a feature that wasn't available in 1.5.2. I
>>frankly don't remember exactly what the issue was, I just remember
>>that Mike Sussman or Ben discovered that something didn't work with
>>pre-2.0 Pythons, did a little research, and found out that we'd made
>>use of a feature new in 2.0. In other words, 2.0 turned out to be
>>backward-compatible but not forward-compatible with 1.5.2... as is
>>usually the case with languages, I suppose.
one more thing that python 2 adds is list comprehensions and some other
syntatic sugar for function calling, which apparently is the problem in the
test suites. it only affects one file actions.py and it looks easy enough to
fix.
>>SOLUTION:
i'm attaching the diff of the changes to make it work with python 1.5.2
diff actions.py oldactions.py > actions.diff
cheers
kapil
Re: Two problems w/ subversion
Posted by "ibej <brane@xbc.nu>" <Branko>.
Joe Orton wrote:
>On Wed, Jul 11, 2001 at 01:20:10AM +0200, Branko Hibej wrote:
>
>>Aargh. What if I only have python2 installed? And gen-make.py refers to
>>python. I'm screwed before I even get to configure.
>>
>>Well, bad luck. My hack goes back. :-)
>>
>
>You could install both again, then gen-make works, and configure picks
>up python2 correctly?
>
>Works for me...
>
Sure. I'm not implying we should try to fix gen-make.py, or anything.
Rather, package maintainers should behave nicely, to do this.
$ ls -l python*
.... python -> python2
.... python1-> python1.5.3
.... python1.5.3*
.... python2 -> python2.0
.... python2.0*
Emacs does that, Tcl does that, I don't see why Python couldn't do that.
I don't imagine there are many packages out there that require Python
1.5, but can't work quite well with Python 2.x.
Anyway, this is off-topic here.
Brane
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Joe Orton <jo...@btconnect.com>.
On Wed, Jul 11, 2001 at 01:20:10AM +0200, Branko Hibej wrote:
> Aargh. What if I only have python2 installed? And gen-make.py refers to
> python. I'm screwed before I even get to configure.
>
> Well, bad luck. My hack goes back. :-)
You could install both again, then gen-make works, and configure picks
up python2 correctly?
Works for me...
joe
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Branko Čibej <br...@xbc.nu>.
Branko �ibej wrote:
> Joe Orton wrote:
>
>> On Tue, Jul 10, 2001 at 05:14:34PM -0500, Karl Fogel wrote:
>> ...
>>
>>> However, since some OS's are apparently installing Python 2.0 as
>>> "python2", if we can make Subversion check for "python2" during the
>>> configure-time search for a useable Python, and then substitute
>>> "python2" in the appropriate places if it finds it, then things should
>>> work correctly for more people.
>>>
>>> Any volunteers, preferably someone with a box that has both Pythons
>>> installed?
>>>
>>
>> I've done it, commits coming up.
>>
>
> Heh. I'd installed python2 on Debian, threw away the old pything and
> made a sumlink python->python2 directly in /usr/bin. Ugly, but it
> works. It's nice that I can get rid of this hack now. :-)
>
> Brane
>
Aargh. What if I only have python2 installed? And gen-make.py refers to
python. I'm screwed before I even get to configure.
Well, bad luck. My hack goes back. :-)
Brane
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Branko Čibej <br...@xbc.nu>.
Joe Orton wrote:
>On Tue, Jul 10, 2001 at 05:14:34PM -0500, Karl Fogel wrote:
>...
>
>>However, since some OS's are apparently installing Python 2.0 as
>>"python2", if we can make Subversion check for "python2" during the
>>configure-time search for a useable Python, and then substitute
>>"python2" in the appropriate places if it finds it, then things should
>>work correctly for more people.
>>
>>Any volunteers, preferably someone with a box that has both Pythons
>>installed?
>>
>
>I've done it, commits coming up.
>
Heh. I'd installed python2 on Debian, threw away the old pything and
made a sumlink python->python2 directly in /usr/bin. Ugly, but it works.
It's nice that I can get rid of this hack now. :-)
Brane
--
Brane �ibej
home: <br...@xbc.nu> http://www.xbc.nu/brane/
work: <br...@hermes.si> http://www.hermes-softlab.com/
ACM : <br...@acm.org> http://www.acm.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Karl Fogel <kf...@collab.net>.
Joe Orton <jo...@btconnect.com> writes:
> > However, since some OS's are apparently installing Python 2.0 as
> > "python2", if we can make Subversion check for "python2" during the
> > configure-time search for a useable Python, and then substitute
> > "python2" in the appropriate places if it finds it, then things should
> > work correctly for more people.
> >
> > Any volunteers, preferably someone with a box that has both Pythons
> > installed?
>
> I've done it, commits coming up.
Thanks! (That was fast!)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Joe Orton <jo...@btconnect.com>.
On Tue, Jul 10, 2001 at 05:14:34PM -0500, Karl Fogel wrote:
...
> However, since some OS's are apparently installing Python 2.0 as
> "python2", if we can make Subversion check for "python2" during the
> configure-time search for a useable Python, and then substitute
> "python2" in the appropriate places if it finds it, then things should
> work correctly for more people.
>
> Any volunteers, preferably someone with a box that has both Pythons
> installed?
I've done it, commits coming up.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Karl Fogel <kf...@collab.net>.
Thanks for the summary, Trent.
Some history here:
Subversion does require Python 2.0 for a reason: there's some code in
the test suite that uses a feature that wasn't available in 1.5.2. I
frankly don't remember exactly what the issue was, I just remember
that Mike Sussman or Ben discovered that something didn't work with
pre-2.0 Pythons, did a little research, and found out that we'd made
use of a feature new in 2.0. In other words, 2.0 turned out to be
backward-compatible but not forward-compatible with 1.5.2... as is
usually the case with languages, I suppose.
SOLUTION:
I doubt it's worth it to make the test suite run with 1.5.2 and 2.0,
since the 2.x series will probably be widespread fairly soon.
However, since some OS's are apparently installing Python 2.0 as
"python2", if we can make Subversion check for "python2" during the
configure-time search for a useable Python, and then substitute
"python2" in the appropriate places if it finds it, then things should
work correctly for more people.
Any volunteers, preferably someone with a box that has both Pythons
installed?
-K
Trent Mick <tr...@ActiveState.com> writes:
> The differences between the two are not *so* large (Unicode support was added
> after Python 1.5.2 but I don't think that is an issue here yet). However,
> some syntactic conveniences provided in Python 2.0 are so nice that they are
> widespread. For example,
> - string methods:
> Before you had to do this:
> import string
> print string.split("hello there buddy", " ")
> Now you can do this:
> s = "hello there buddy"
> s.split(" ")
> - augmented assignment: i.e. +=,-=,*=,etc.
>
> c.f. http://amk.ca/python/2.0/
>
>
> Trent
>
> --
> Trent Mick
> TrentM@ActiveState.com
>
> ---------------------------------------------------------------------
> 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: Two problems w/ subversion
Posted by Trent Mick <tr...@ActiveState.com>.
On Tue, Jul 10, 2001 at 11:56:27AM -0700, Brian Behlendorf wrote:
> Are there incompatible differences between python 1 and 2 that mandate the
> difference? Are they significant, or small cleanups like the perl 4 vs 5
> difference?
The differences between the two are not *so* large (Unicode support was added
after Python 1.5.2 but I don't think that is an issue here yet). However,
some syntactic conveniences provided in Python 2.0 are so nice that they are
widespread. For example,
- string methods:
Before you had to do this:
import string
print string.split("hello there buddy", " ")
Now you can do this:
s = "hello there buddy"
s.split(" ")
- augmented assignment: i.e. +=,-=,*=,etc.
c.f. http://amk.ca/python/2.0/
Trent
--
Trent Mick
TrentM@ActiveState.com
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Brian Behlendorf <br...@collab.net>.
On 10 Jul 2001, Bryan O'Sullivan wrote:
> c> I'm not sure I'd call this a problem, really. I mean, isn't the
> c> convention to have "python" a symlink to the latest greatest thing,
> c> and to invoke *older* versions by hand (with "python-1.5.2" or
> c> something)?
>
> Not under Linux. All Red Hat based distributions that I know of have
> "python" as 1.5.2, as does Debian, and all use "python2" as the more
> recent version.
Are there incompatible differences between python 1 and 2 that mandate the
difference? Are they significant, or small cleanups like the perl 4 vs 5
difference?
Brian
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Bryan O'Sullivan <bo...@serpentine.com>.
c> I'm not sure I'd call this a problem, really. I mean, isn't the
c> convention to have "python" a symlink to the latest greatest thing,
c> and to invoke *older* versions by hand (with "python-1.5.2" or
c> something)?
Not under Linux. All Red Hat based distributions that I know of have
"python" as 1.5.2, as does Debian, and all use "python2" as the more
recent version.
<b
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by cm...@collab.net.
Jon Trowbridge <tr...@ximian.com> writes:
> (1) A minor problem.
>
> The test scripts (basic-tests.py in particular) seem to require python
> 2.0. On my machine (running debian Woody), "python" invokes python
> 1.5.2 and "python2" runs 2.0. Thus things like basic-tests.py won't run
> for me unless I manually run them from the command line with "python2
> ./basic-tests.py".
I'm not sure I'd call this a problem, really. I mean, isn't the
convention to have "python" a symlink to the latest greatest thing,
and to invoke *older* versions by hand (with "python-1.5.2" or
something)?
> (2) A larger problem
>
> svnadmin segfaults when I try to create a repository.
I know this was happening a couple of weeks ago, but that bug has
since been fixed. I use svnadmin daily here to create repositories,
and am not seeing the problem. Any chance that your binary is out of
date?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Two problems w/ subversion
Posted by Karl Fogel <kf...@collab.net>.
Jon Trowbridge <tr...@ximian.com> writes:
> (1) A minor problem.
>
> The test scripts (basic-tests.py in particular) seem to require python
> 2.0. On my machine (running debian Woody), "python" invokes python
> 1.5.2 and "python2" runs 2.0. Thus things like basic-tests.py won't run
> for me unless I manually run them from the command line with "python2
> ./basic-tests.py".
Hmmm. Not really sure what a good fix for this would be; unless
there's some way to make autoconf discover python 2.0 on your machine
under the name "python2"?
I know you probably hate hearing this, but running them manually until
you upgrade python isn't the worst thing in the world (not as serious
as the autoconf 2.13 vs 2.50 issue, for example). :-) Unlike autoconf
and similar tools, languages don't generally have serious
backward-incompatibilities, so asking people to upgrade their pythons
seems like a reasonable thing. Upgrading to python2 should break
nothing, and make it possible to run more code.
-K
> (2) A larger problem
>
> svnadmin segfaults when I try to create a repository. Here is a
> backtrace
Working on it right now with Mike Pilato <cm...@collab.net>.
Thanks for the report!
-K
> (gdb) run create /home/trow/svn-test-repos
> Starting program: /home/trow/bin/svnadmin create
> /home/trow/svn-test-repos
>
> Program received signal SIGSEGV, Segmentation fault.
> 0x0 in ?? ()
> (gdb) bt
> #0 0x0 in ?? ()
> #1 0x804ad77 in cleanup_fs (fs=0x80736a4) at
> subversion/libsvn_fs/fs.c:143
> #2 0x804b572 in svn_fs_create_berkeley (fs=0x80736a4,
> path=0x7ffffa22 "/home/trow/svn-test-repos")
> at subversion/libsvn_fs/fs.c:436
> #3 0x804a5f3 in main (argc=3, argv=0x7ffff8c4)
> at subversion/svnadmin/main.c:159
>
> I'm using the db3 packages provided by debian.
>
> -JT
>
>
>
>
> ---------------------------------------------------------------------
> 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