You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Moisei Rabinovich <mo...@gmail.com> on 2006/10/03 02:20:20 UTC
subversion 1.4.0 svn checkout crash
Hi,
I just upgraded to 1.4.0 and now I have a crash when I try to checkout my
working copy. the "crash" means standard windows error pop-up and working
copy remains locked.
It consistently happens in the middle of the checkout.
The scenario is very simple - checkout fresh copy via http [svn checkout "
http://myserver/svn/trunk" "S:\working-copy"]
S is a subst drive.
The server is continuous integration server and svn command is executed as a
part of the build script, as an "execute" ant task.
NOTE: I can *NOT* reproduce this crash when running the same command from
the command line window.
NOTE: It was perfectly working with 1.3.2 for ~ 1 year.
This is the exactly ant task I use:
<target name="test-svn-checkout">
<exec executable="svn.exe" failonerror="true">
<arg value="checkout"/>
<arg value="--non-interactive"/>
<arg value="--username"/>
<arg value="DOMAIN\user"/>
<arg value="--password"/>
<arg value="password"/>
<arg value="http://example/repo"/>
<arg value="S:\working copy"/>
</exec>
</target>
my server env:
Intel Xeon CPU 3.00 GHz 1.00 GB of RAM
Microsoft Windows Server 2003 Standard Edition Service Pack 1
Apache 2.0.54
svn, version 1.4.0 (r21228) compiled Sep 11 2006, 17:46:59 [installed from
the svn-1.4.0-setup.exe]
my client env:
Intel Xeon CPU2.80 GHz 1.00 GB of RAM
Microsoft Windows Server 2003 Standard Edition Service Pack 1
svn, version 1.4.0 (r21228) compiled Sep 11 2006, 17:46:59 [installed from
the svn-1.4.0-setup.exe]
ant 1.6.2
--
Best Regards,
Moisei
Re: subversion 1.4.0 svn checkout crash
Posted by Lieven Govaerts <sv...@mobsol.be>.
Moisei Rabinovich wrote:
> For svn windows dev: IMHO under any condition the pop up error dialog
> is not acceptable - it freezes the entire process
> without report an error back to the caller application that is the
> worse behavior one can imagine for any automation system
> There are some techniques to prevent windows application to raise such
> a dialogs, I would consider to use it for svn application.
Which dialog is that? The 'report this error to Microsoft?' dialog?
Does Windows Server 2003 have Dr. Watson installed? If yes, it probably
has stored logs of the previous svn crashes already. More info here:
http://support.microsoft.com/kb/308538 .
Lieven.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by "D.J. Heap" <dj...@gmail.com>.
On 10/3/06, Moisei Rabinovich <mo...@gmail.com> wrote:
[snip]
> For svn windows dev: IMHO under any condition the pop up error dialog is not
> acceptable - it freezes the entire process
> without report an error back to the caller application that is the worse
> behavior one can imagine for any automation system
> There are some techniques to prevent windows application to raise such a
> dialogs, I would consider to use it for svn application.
>
I agree -- can you work up a patch to implement this?
DJ
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Lieven Govaerts <sv...@mobsol.be>.
Moisei Rabinovich wrote:
> For svn windows dev: IMHO under any condition the pop up error dialog
> is not acceptable - it freezes the entire process
> without report an error back to the caller application that is the
> worse behavior one can imagine for any automation system
> There are some techniques to prevent windows application to raise such
> a dialogs, I would consider to use it for svn application.
Which dialog is that? The 'report this error to Microsoft?' dialog?
Does Windows Server 2003 have Dr. Watson installed? If yes, it probably
has stored logs of the previous svn crashes already. More info here:
http://support.microsoft.com/kb/308538 .
Lieven.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by "D.J. Heap" <dj...@gmail.com>.
On 10/3/06, Moisei Rabinovich <mo...@gmail.com> wrote:
[snip]
> For svn windows dev: IMHO under any condition the pop up error dialog is not
> acceptable - it freezes the entire process
> without report an error back to the caller application that is the worse
> behavior one can imagine for any automation system
> There are some techniques to prevent windows application to raise such a
> dialogs, I would consider to use it for svn application.
>
I agree -- can you work up a patch to implement this?
DJ
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Moisei Rabinovich <mo...@gmail.com>.
> Can you get a stack trace from svn.exe or Ant? Or is there anything
> akin to strace that you can run when invoking Ant?
I wish I knew how to do it. This is what all this thread about: give svn
crashes on windows - how to get enough
information to report the crash or what troubleshooting to execute to
understand the problem.
Currently the only relevant answer I got is to run it via batch file and see
whether it helps but even if it helps it does not
eliminate the problem but just hides it. In the meanwhile, I downgraded the
client machine back to svn 1.3.2 and it works fine.
I will post a result of svn batch as soon as I have a time to reproduce it.
Still any suggestion is strongly appreciated even it is not urgent problem
for me anymore,
I believe it is better to resolve it before it becomes catastrophic for
somebody else.
For svn windows dev: IMHO under any condition the pop up error dialog is not
acceptable - it freezes the entire process
without report an error back to the caller application that is the worse
behavior one can imagine for any automation system
There are some techniques to prevent windows application to raise such a
dialogs, I would consider to use it for svn application.
For svn dev: in fact the only "suspicious" aspect of running svn via ant
task that I can think about is
memory limitation that ant may put on forked process... Do you have any
validation on svn binaries under limited memory scope?
May be I could try to run it on my env to see whether it is reproduced.
--
Best Regards,
Moisei
Re: subversion 1.4.0 svn checkout crash
Posted by Moisei Rabinovich <mo...@gmail.com>.
> Can you get a stack trace from svn.exe or Ant? Or is there anything
> akin to strace that you can run when invoking Ant?
I wish I knew how to do it. This is what all this thread about: give svn
crashes on windows - how to get enough
information to report the crash or what troubleshooting to execute to
understand the problem.
Currently the only relevant answer I got is to run it via batch file and see
whether it helps but even if it helps it does not
eliminate the problem but just hides it. In the meanwhile, I downgraded the
client machine back to svn 1.3.2 and it works fine.
I will post a result of svn batch as soon as I have a time to reproduce it.
Still any suggestion is strongly appreciated even it is not urgent problem
for me anymore,
I believe it is better to resolve it before it becomes catastrophic for
somebody else.
For svn windows dev: IMHO under any condition the pop up error dialog is not
acceptable - it freezes the entire process
without report an error back to the caller application that is the worse
behavior one can imagine for any automation system
There are some techniques to prevent windows application to raise such a
dialogs, I would consider to use it for svn application.
For svn dev: in fact the only "suspicious" aspect of running svn via ant
task that I can think about is
memory limitation that ant may put on forked process... Do you have any
validation on svn binaries under limited memory scope?
May be I could try to run it on my env to see whether it is reproduced.
--
Best Regards,
Moisei
Re: subversion 1.4.0 svn checkout crash
Posted by Daniel Rall <dl...@collab.net>.
On Tue, 03 Oct 2006, Moisei Rabinovich wrote:
...
> I just upgraded to 1.4.0 and now I have a crash when I try to checkout my
> working copy. the "crash" means standard windows error pop-up and working
> copy remains locked.
> It consistently happens in the middle of the checkout.
> The scenario is very simple - checkout fresh copy via http [svn checkout "
> http://myserver/svn/trunk" "S:\working-copy"]
> S is a subst drive.
> The server is continuous integration server and svn command is executed as a
> part of the build script, as an "execute" ant task.
>
> NOTE: I can *NOT* reproduce this crash when running the same command from
> the command line window.
This sounds like a bad interaction between Ant and the svn.exe client
(as you seem to suspect, too).
...
> <target name="test-svn-checkout">
> <exec executable="svn.exe" failonerror="true">
> <arg value="checkout"/>
> <arg value="--non-interactive"/>
> <arg value="--username"/>
> <arg value="DOMAIN\user"/>
> <arg value="--password"/>
> <arg value="password"/>
> <arg value="http://example/repo"/>
> <arg value="S:\working copy"/>
> </exec>
> </target>
...
> my client env:
> Intel Xeon CPU2.80 GHz 1.00 GB of RAM
> Microsoft Windows Server 2003 Standard Edition Service Pack 1
> svn, version 1.4.0 (r21228) compiled Sep 11 2006, 17:46:59 [installed from
> the svn-1.4.0-setup.exe]
> ant 1.6.2
Can you get a stack trace from svn.exe or Ant? Or is there anything
akin to strace that you can run when invoking Ant?
Re: subversion 1.4.0 svn checkout crash
Posted by Samay <ge...@hotmail.com>.
Moisei,
is error message similar to one here @ http://svn.haxx.se/users/archive-2006-09/1362.shtml
"The instruction at "0x00000000" referenced memory at "0x000000000". The memory could not be "read".
If yes, i guess points to neon 0.26.0 changes over 0.25.x.
regards
S.
----- Original Message -----
From: Moisei Rabinovich
To: John Peacock
Cc: Ben Collins-Sussman ; dev@subversion.tigris.org
Sent: Tuesday, October 03, 2006 1:12 PM
Subject: Re: subversion 1.4.0 svn checkout crash
OP is probably me... just fyi - OP is doing his best to provide as much info as possible.
Upgrade to 1.4.0 is *NOT* coincidental with some other change in the environment. I do not change two things at time.
I downgraded the client back to 1.3.2 and I do not have a crash anymore.
The fact 1.3.2 works well and the fact 1.4.0 works well from the command line, probably eliminates the "antivirus theory".
I also tried to use ant 1.7.beta2 - with the exactly same result.
I also tried to increase memory usage for ant using ANT_OPTS - with the exactly same result.
> 1) the OP didn't tell us what, if anything, the Windows error popup said;
I will post the screenshot next time I reproduce it. This is usual windows error message telling "bad memory access on ... address"
> 2) the OP didn't tell us whether the failure is always on the same file or on
> different files;
On different files but in the same folder.
> 3) the OP didn't tell us whether the subst'd drive is a local or networked drive;
subst always mean local. network drive called "map-drive"
> 4) the OP didn't tell us whether the client box was running any antivirus
> software (or other software that hooks into the filesystem);
yes, it runs semantec antivirus.
> 5) the OP didn't tell us whether this only happened only when two or more
> commits were made in quick succession, nor how large the working copy is.
only one commit at time. no other activity on the server and on the client.
Re: subversion 1.4.0 svn checkout crash
Posted by Matt Doran <ma...@papercut.biz>.
Moisei Rabinovich wrote:
> OP is probably me... just fyi - OP is doing his best to provide as
> much info as possible.
>
> Upgrade to 1.4.0 is *NOT* coincidental with some other change in the
> environment. I do not change two things at time.
> I downgraded the client back to 1.3.2 and I do not have a crash anymore.
> The fact 1.3.2 works well and the fact 1.4.0 works well from the
> command line, probably eliminates the "antivirus theory".
> I also tried to use ant 1.7.beta2 - with the exactly same result.
> I also tried to increase memory usage for ant using ANT_OPTS - with
> the exactly same result.
>
>
Hi there,
Ant can do some pretty serious voodoo to try to exec commands. It does
this to work around the various quirks with how Java execs programs on
Windows. But the command it actually executes depends on the version of
Java and Windows. For example, if it needs to changes the current
working directory, I think it executes something like this:
cmd.exe /c cd /d <new-working-dir> && <original command>
It might be worth trying to replicate what Ant is executing. Or maybe
trying to use Ant to exec a batch file that runs your svn command and
see if you get the same behaviour.
Regards,
Matt Doran
PaperCut Software Pty. Ltd.
Phone: +61 (3) 9807 5767
E-mail: matt.doran@papercut.biz
Profile: http://papercut.biz/company.htm#matt
Blog: http://papercut.biz/blog
Re: subversion 1.4.0 svn checkout crash
Posted by Moisei Rabinovich <mo...@gmail.com>.
OP is probably me... just fyi - OP is doing his best to provide as much info
as possible.
Upgrade to 1.4.0 is *NOT* coincidental with some other change in the
environment. I do not change two things at time.
I downgraded the client back to 1.3.2 and I do not have a crash anymore.
The fact 1.3.2 works well and the fact 1.4.0 works well from the command
line, probably eliminates the "antivirus theory".
I also tried to use ant 1.7.beta2 - with the exactly same result.
I also tried to increase memory usage for ant using ANT_OPTS - with the
exactly same result.
> 1) the OP didn't tell us what, if anything, the Windows error popup said;
I will post the screenshot next time I reproduce it. This is usual windows
error message telling "bad memory access on ... address"
> 2) the OP didn't tell us whether the failure is always on the same file or
on
> different files;
On different files but in the same folder.
> 3) the OP didn't tell us whether the subst'd drive is a local or networked
drive;
subst always mean local. network drive called "map-drive"
> 4) the OP didn't tell us whether the client box was running any antivirus
> software (or other software that hooks into the filesystem);
yes, it runs semantec antivirus.
> 5) the OP didn't tell us whether this only happened only when two or more
> commits were made in quick succession, nor how large the working copy is.
only one commit at time. no other activity on the server and on the client.
Re: subversion 1.4.0 svn checkout crash
Posted by John Peacock <jp...@rowman.com>.
Ben Collins-Sussman wrote:
> Win32 folks: how do we collect a backtrace in situations like this?
>
Several things came to mind when I read the original message:
1) the OP didn't tell us what, if anything, the Windows error popup said;
2) the OP didn't tell us whether the failure is always on the same file or on
different files;
3) the OP didn't tell us whether the subst'd drive is a local or networked drive;
4) the OP didn't tell us whether the client box was running any antivirus
software (or other software that hooks into the filesystem);
5) the OP didn't tell us whether this only happened only when two or more
commits were made in quick succession, nor how large the working copy is.
My gut instinct says to obtain the specifics of the machine(s) involved in much
more detail before spending that much time trying to get a backtrace. The
upgrade to 1.4.0 may be coincidental with some other change in the environment.
John
--
John Peacock
Director of Information Research and Technology
Rowman & Littlefield Publishing Group
4501 Forbes Blvd
Suite H
Lanham, MD 20706
301-459-3366 x.5010
fax 301-429-5747
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by "D.J. Heap" <dj...@gmail.com>.
On 11/18/06, Lieven Govaerts <sv...@mobsol.be> wrote:
> D.J. Heap wrote:
> > Based on the long thread, it sounds like you've tweaked the source a
> > bit, Stefan? Do you have the updated source available so I could
> > build it with VC6 as you suggested to avoid having to install another
> > C runtime?
> At the summit, Ivan and I started looking at the crash handler code
> examples + the TSVN version. We concluded that it'd be better if we code
> this from scratch, with just the things we need.
> That was a month ago and we didn't really follow up, but I have no
> problem looking at it the coming days.
>
> Is there any special reason why we want to put this code in a separate
> dll? It seems to work fine with both the setup and crash handler code in
> the executable.
>
> Lieven
>
No special reason except that it's already done and working. Either
way is fine with me.
DJ
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Malcolm Rowe <ma...@farside.org.uk>.
On Sat, Nov 18, 2006 at 07:09:09PM +0100, Stefan Küng wrote:
> >Is there any special reason why we want to put this code in a separate
> >dll? It seems to work fine with both the setup and crash handler code in
> >the executable.
>
> If you put it into the executable, then it won't work. After all: it has
> to catch a segfault, and a segfault will *stop* the application. That's
> why it is implemented as a dll.
>
Er, what? The DLL is loaded into the address space of the process -
there's nothing special about whether it's a DLL or not, it just calls
SetUnhandledExceptionFilter() to set a handler (in the same way that you
can call signal() to handle SIGSEGV before terminating).
> But are you sure you want to start from scratch with a crashhandler? You
> could also start using airbag:
> http://code.google.com/p/airbag/
>
Airbag looks like it'll be good, but it was pretty incomplete last time
I checked. It also seemed more focussed around sending the crash
reports to a server.
Regards,
Malcolm
Re: subversion 1.4.0 svn checkout crash
Posted by Stefan Küng <to...@gmail.com>.
Lieven Govaerts wrote:
> D.J. Heap wrote:
>> Based on the long thread, it sounds like you've tweaked the source a
>> bit, Stefan? Do you have the updated source available so I could
>> build it with VC6 as you suggested to avoid having to install another
>> C runtime?
> At the summit, Ivan and I started looking at the crash handler code
> examples + the TSVN version. We concluded that it'd be better if we code
> this from scratch, with just the things we need.
> That was a month ago and we didn't really follow up, but I have no
> problem looking at it the coming days.
Well, I certainly have nothing against you starting from scratch.
But I have to say, too often people on this list reject a solution
because they want something 'better' just because they want it perfect.
I'd rather have a *working* solution than something perfect which might
never get done.
> Is there any special reason why we want to put this code in a separate
> dll? It seems to work fine with both the setup and crash handler code in
> the executable.
If you put it into the executable, then it won't work. After all: it has
to catch a segfault, and a segfault will *stop* the application. That's
why it is implemented as a dll.
But are you sure you want to start from scratch with a crashhandler? You
could also start using airbag:
http://code.google.com/p/airbag/
Stefan
--
___
oo // \\ "De Chelonian Mobile"
(_,\/ \_/ \ TortoiseSVN
\ \_/_\_/> The coolest Interface to (Sub)Version Control
/_/ \_\ http://tortoisesvn.net
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Lieven Govaerts <sv...@mobsol.be>.
D.J. Heap wrote:
> Based on the long thread, it sounds like you've tweaked the source a
> bit, Stefan? Do you have the updated source available so I could
> build it with VC6 as you suggested to avoid having to install another
> C runtime?
At the summit, Ivan and I started looking at the crash handler code
examples + the TSVN version. We concluded that it'd be better if we code
this from scratch, with just the things we need.
That was a month ago and we didn't really follow up, but I have no
problem looking at it the coming days.
Is there any special reason why we want to put this code in a separate
dll? It seems to work fine with both the setup and crash handler code in
the executable.
Lieven
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Stefan Küng <to...@gmail.com>.
D.J. Heap wrote:
> On 10/6/06, Stefan Küng <to...@gmail.com> wrote:
> [snip]
>> > Can this be tweaked to write out the dump and leave a pointer to it in
>> > the event log as Brane suggested? Also, what is the license for the
>> > crashreport dll? Could we just include the code in Windows builds?
>>
>> The revised patch at the very end of the mailing thread does exactly
>> that: it prints out an error with the path where the dump file was
>> written to. And a text to 'please send the dump to
>> mailaddress@subversion'.
>>
>> Stefan
>
>
> I've investigated this further and found the license for the crash
> handler DLL (license.txt in the source directory). It looks fine to
> me:
[snip]
>
> Based on the long thread, it sounds like you've tweaked the source a
> bit, Stefan? Do you have the updated source available so I could
> build it with VC6 as you suggested to avoid having to install another
> C runtime?
The source is in the TSVN repository:
http://tortoisesvn.tigris.org/svn/tortoisesvn/trunk/src/crashrpt/
All the tweak necessary to just save the dumpfile without any UI are
already committed.
You might need the WTL library (I haven't tweaked the compile time
settings to not include the UI stuff, but it runs also without any UI):
http://sourceforge.net/projects/wtl/
then you should be able to compile it with VC6 as well.
Stefan
--
___
oo // \\ "De Chelonian Mobile"
(_,\/ \_/ \ TortoiseSVN
\ \_/_\_/> The coolest Interface to (Sub)Version Control
/_/ \_\ http://tortoisesvn.net
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by "D.J. Heap" <dj...@gmail.com>.
On 10/6/06, Stefan Küng <to...@gmail.com> wrote:
[snip]
> > Can this be tweaked to write out the dump and leave a pointer to it in
> > the event log as Brane suggested? Also, what is the license for the
> > crashreport dll? Could we just include the code in Windows builds?
>
> The revised patch at the very end of the mailing thread does exactly
> that: it prints out an error with the path where the dump file was
> written to. And a text to 'please send the dump to
> mailaddress@subversion'.
>
> Stefan
I've investigated this further and found the license for the crash
handler DLL (license.txt in the source directory). It looks fine to
me:
Copyright (c) 2003 Michael Carruth
This software is provided 'as-is', without any express or implied
warranty. In no event will
the authors be held liable for any damages arising from the use of
this software.
Permission is granted to anyone to use this software for any purpose,
including commercial
applications, and to alter it and redistribute it freely, subject to
the following
restrictions:
1. The origin of this software must not be misrepresented; you
must not claim that you
wrote the original software. If you use this software in a
product, an acknowledgment in
the product documentation would be appreciated but is not required.
2. Altered source versions must be plainly marked as such, and
must not be misrepresented
as being the original software.
3. This notice may not be removed or altered from any source distribution.
Based on the long thread, it sounds like you've tweaked the source a
bit, Stefan? Do you have the updated source available so I could
build it with VC6 as you suggested to avoid having to install another
C runtime?
DJ
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Stefan Küng <to...@gmail.com>.
On 10/6/06, D.J. Heap <dj...@gmail.com> wrote:
> On 10/3/06, Stefan Küng <to...@gmail.com> wrote:
> > Ben Collins-Sussman wrote:
> > > Win32 folks: how do we collect a backtrace in situations like this?
> >
> > must - resist - yelling - told - you - so
> >
> > Issue #1628 (http://subversion.tigris.org/issues/show_bug.cgi?id=1628)
> > is about implementing a crash handler for windows.
> > I've sent in a patch for this, revised version (after some discussions)
> > is still here:
> > http://svn.haxx.se/dev/archive-2005-07/1069.shtml
>
>
> I'd like to look at this further, but I think popping up a box on
> commandline programs is very bad behavior as has been mentioned before
> -- automated processes may not even be running in a visible
> winstation, so the process will be hung forever with no one even able
> to see the dialog.
>
> Can this be tweaked to write out the dump and leave a pointer to it in
> the event log as Brane suggested? Also, what is the license for the
> crashreport dll? Could we just include the code in Windows builds?
The revised patch at the very end of the mailing thread does exactly
that: it prints out an error with the path where the dump file was
written to. And a text to 'please send the dump to
mailaddress@subversion'.
Stefan
--
___
oo // \\ "De Chelonian Mobile"
(_,\/ \_/ \ TortoiseSVN
\ \_/_\_/> The coolest Interface to (Sub)Version Control
/_/ \_\ http://tortoisesvn.net
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by "D.J. Heap" <dj...@gmail.com>.
On 10/3/06, Stefan Küng <to...@gmail.com> wrote:
> Ben Collins-Sussman wrote:
> > Win32 folks: how do we collect a backtrace in situations like this?
>
> must - resist - yelling - told - you - so
>
> Issue #1628 (http://subversion.tigris.org/issues/show_bug.cgi?id=1628)
> is about implementing a crash handler for windows.
> I've sent in a patch for this, revised version (after some discussions)
> is still here:
> http://svn.haxx.se/dev/archive-2005-07/1069.shtml
I'd like to look at this further, but I think popping up a box on
commandline programs is very bad behavior as has been mentioned before
-- automated processes may not even be running in a visible
winstation, so the process will be hung forever with no one even able
to see the dialog.
Can this be tweaked to write out the dump and leave a pointer to it in
the event log as Brane suggested? Also, what is the license for the
crashreport dll? Could we just include the code in Windows builds?
DJ
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Stefan Küng <to...@gmail.com>.
Ben Collins-Sussman wrote:
> Win32 folks: how do we collect a backtrace in situations like this?
must - resist - yelling - told - you - so
Issue #1628 (http://subversion.tigris.org/issues/show_bug.cgi?id=1628)
is about implementing a crash handler for windows.
I've sent in a patch for this, revised version (after some discussions)
is still here:
http://svn.haxx.se/dev/archive-2005-07/1069.shtml
Even though my patch was working well, did not interfere with Subversion
operations at all it never got committed.
I'll check the crashreports I got for TSVN 1.4.0 now - maybe I can find
one which corresponds to this crash here and provide a stacktrace.
Stefan
--
___
oo // \\ "De Chelonian Mobile"
(_,\/ \_/ \ TortoiseSVN
\ \_/_\_/> The coolest Interface to (Sub)Version Control
/_/ \_\ http://tortoisesvn.net
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Ben Collins-Sussman <su...@red-bean.com>.
Win32 folks: how do we collect a backtrace in situations like this?
On 10/2/06, Moisei Rabinovich <mo...@gmail.com> wrote:
> Hi,
> I just upgraded to 1.4.0 and now I have a crash when I try to checkout my
> working copy. the "crash" means standard windows error pop-up and working
> copy remains locked.
> It consistently happens in the middle of the checkout.
> The scenario is very simple - checkout fresh copy via http [svn checkout
> "http://myserver/svn/trunk" "S:\working-copy"]
> S is a subst drive.
> The server is continuous integration server and svn command is executed as a
> part of the build script, as an "execute" ant task.
>
> NOTE: I can *NOT* reproduce this crash when running the same command from
> the command line window.
>
> NOTE: It was perfectly working with 1.3.2 for ~ 1 year.
>
> This is the exactly ant task I use:
>
> <target name="test-svn-checkout">
> <exec executable="svn.exe" failonerror="true">
> <arg value="checkout"/>
> <arg value="--non-interactive"/>
> <arg value="--username"/>
> <arg value="DOMAIN\user"/>
> <arg value="--password"/>
> <arg value="password"/>
> <arg value="http://example/repo"/>
> <arg value="S:\working copy"/>
> </exec>
> </target>
>
> my server env:
> Intel Xeon CPU 3.00 GHz 1.00 GB of RAM
> Microsoft Windows Server 2003 Standard Edition Service Pack 1
> Apache 2.0.54
> svn, version 1.4.0 (r21228) compiled Sep 11 2006, 17:46:59 [installed from
> the svn-1.4.0-setup.exe ]
>
> my client env:
> Intel Xeon CPU2.80 GHz 1.00 GB of RAM
> Microsoft Windows Server 2003 Standard Edition Service Pack 1
> svn, version 1.4.0 (r21228) compiled Sep 11 2006, 17:46:59 [installed from
> the svn-1.4.0-setup.exe ]
> ant 1.6.2
>
>
> --
> Best Regards,
> Moisei
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: subversion 1.4.0 svn checkout crash
Posted by Daniel Rall <dl...@collab.net>.
On Tue, 03 Oct 2006, Moisei Rabinovich wrote:
...
> I just upgraded to 1.4.0 and now I have a crash when I try to checkout my
> working copy. the "crash" means standard windows error pop-up and working
> copy remains locked.
> It consistently happens in the middle of the checkout.
> The scenario is very simple - checkout fresh copy via http [svn checkout "
> http://myserver/svn/trunk" "S:\working-copy"]
> S is a subst drive.
> The server is continuous integration server and svn command is executed as a
> part of the build script, as an "execute" ant task.
>
> NOTE: I can *NOT* reproduce this crash when running the same command from
> the command line window.
This sounds like a bad interaction between Ant and the svn.exe client
(as you seem to suspect, too).
...
> <target name="test-svn-checkout">
> <exec executable="svn.exe" failonerror="true">
> <arg value="checkout"/>
> <arg value="--non-interactive"/>
> <arg value="--username"/>
> <arg value="DOMAIN\user"/>
> <arg value="--password"/>
> <arg value="password"/>
> <arg value="http://example/repo"/>
> <arg value="S:\working copy"/>
> </exec>
> </target>
...
> my client env:
> Intel Xeon CPU2.80 GHz 1.00 GB of RAM
> Microsoft Windows Server 2003 Standard Edition Service Pack 1
> svn, version 1.4.0 (r21228) compiled Sep 11 2006, 17:46:59 [installed from
> the svn-1.4.0-setup.exe]
> ant 1.6.2
Can you get a stack trace from svn.exe or Ant? Or is there anything
akin to strace that you can run when invoking Ant?
Re: subversion 1.4.0 svn checkout crash
Posted by Pedro Lamarão <pe...@intersix.com.br>.
Moisei Rabinovich escreveu:
> Hi,
> I just upgraded to 1.4.0 and now I have a crash when I try to checkout
> my working copy. the "crash" means standard windows error pop-up and
> working copy remains locked.
If there is an error popup then what is the error message?
--
Pedro Lamarão
Desenvolvimento
Intersix Technologies S.A.
SP: (55 11 3803-9300)
RJ: (55 21 3852-3240)
www.intersix.com.br
Your Security is our Business
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org