You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by kf...@collab.net on 2004/01/15 23:45:59 UTC
remember to delete subversion/libsvn_subr/getdate.c
If you're building from a working copy of the 1.0-stabilization branch
or trunk, you might need to do
C:> del subversion\libsvn_subr\getdate.c
or
$ rm subversion/libsvn_subr/getdate.c
after your next update. Although getdate.c is not versioned, the
build system will still see it, say "Aha, it's a C file", and include
it in the build. But you don't want to compile it anymore, because
the date parser has been rewritten, and getdate.c is no longer used.
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svnserve hang
Posted by Pete Gonzalez <pg...@bluel.com>.
At 10:29 PM 1/15/2004, you wrote:
>to figure out why it hangs. Unfortunately, I did not
>compile this version with --enable-maintainer-mode. I
>intend to rebuild the server soon with this option.
>I am wondering if there is anywhere else I can look to
>try to figure out why this occurs?
Try the "strace" tool, which logs all the system calls
by a process. It will tell you at least which C library
function is stalling. It's not perfect, but sometimes it
can be just the battle axe you need.
Cheers,
-Pete
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svnserve hang - SMP questions
Posted by Tobias Ringstrom <to...@ringstrom.mine.nu>.
root wrote:
> --- root <eb...@yahoo.com> wrote:
>
>>--- Tobias Ringstrom <to...@ringstrom.mine.nu>
>>wrote:
>
> Tobias -- I just wanted to follow up with my issue. I
> upgraded to 0.36 and the problem appears to have gone
> away. Now, svn+ssh does not hang in the middle of a
> lengthy checkout.
OK, great!
> I also compiled with --enable-maintaner-mode. I am
> curious what debugging ability this adds? Does it
> simply compile with -g, or will it generate debug
> messages to stdout or some logfile somewhere?
It's compiled with -g and the error messages are more verbose (showing
the filename and line numbers where the errors occurred). I don't know
if it does more than that, but I'm pretty sure that it does not put log
files anywhere.
/Tobias
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svnserve hang - SMP questions
Posted by root <eb...@yahoo.com>.
--- root <eb...@yahoo.com> wrote:
>
> --- Tobias Ringstrom <to...@ringstrom.mine.nu>
> wrote:
Tobias -- I just wanted to follow up with my issue. I
upgraded to 0.36 and the problem appears to have gone
away. Now, svn+ssh does not hang in the middle of a
lengthy checkout.
I also compiled with --enable-maintaner-mode. I am
curious what debugging ability this adds? Does it
simply compile with -g, or will it generate debug
messages to stdout or some logfile somewhere?
Regards,
Davis
__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free web site building tool. Try it!
http://webhosting.yahoo.com/ps/sb/
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svnserve hang - SMP questions
Posted by root <eb...@yahoo.com>.
--- Tobias Ringstrom <to...@ringstrom.mine.nu> wrote:
> root wrote:
> > Hi Tobias,
> >
> > Thanks for the response. I realize that what you
> > indicate below is covered in the FAQ. However, I
> do
> > not believe this to be the case. I am the lone
> > administrator of this box and this repository. I
> can
> > set up a brand new clean repository, and then from
> the
> > client, do a checkout and the same thing occurs --
> it
> > hangs. I can guarantee that there are no other
> > clients hitting the repository or local svn*
> processes
> > hitting the repository at the same time the
> checkout
>
> It's completely OK to have several clients accessing
> the repository at
> the same time of course.
Right - I would think this to be the case..I was
simply trying to rule out the scenario where others'
local/remote accesses were leaving unfinished
transactions.
>
> > is underway. Doing svnadmin recover does not seem
> to
> > fix this problem.
>
> I assume that you killed all svnserve:s, ran
> svnadmin recover, and that
> it completed without error. We need to narrow this
> down a bit further.
> Make sure that your repos is not a network drive,
> and that you do not
> have a permission/umask problem.
Yes, to your question on running svnadmin recover.
This is definitely *not* a network drive.
Permission/umask problem - I've been trying
dilligently to avoid these issues. I have a wrapper
script for svnserve that sets the umask to 0007. The
permissions on the repository are as follows with all
users being members of svn group:
drwsrws--- 2 svn svn 4096 Jan 12
20:46 conf
drwsrws--- 2 svn svn 4096 Jan 12
20:46 dav
drwsrws--- 2 svn svn 118784 Jan 20
09:39 db
-rwsrws--- 1 svn svn 2 Jan 12
20:46 format
drwsrws--- 2 svn svn 4096 Jan 12
20:46 hooks
drwsrws--- 2 svn svn 4096 Jan 12
20:46 locks
-rwsrws--- 1 svn svn 376 Jan 12
20:46 README.txt
>
> Can you checkout using file://? If that works, can
> you run "svnserve -d"
> on the server and access it via svn://?
I can successfully do a file:/// checkout of the
repository - this problem only seems to happen over
the network.
I'll try to manually run svnserve -d tonight and see
if I can checkout successfully via that method.
>
> You should also upgrade to the latest release (0.36)
> of course.
I plan to do this tonight with
--enable-maintainer-mode
>
> > What I am curious about -- is if it is possibly an
> SMP
> > issue. Has anyone run svnserve on SMP boxes
> before?
>
> I doubt that this is an SMP problem, but who can
> know for sure?
>
> > What I see is that a single client doing a
> checkout
> > connects via SSH one time and starts the svnserve
> > process, then a second or so later connects via
> SSH
> > again, and launches *another* svnserve process.
> Is
> > this by design -- such that a single checkout
> requires
> > 2 svnserve processes to be running? If so, is it
> at
>
> Yes.
Just out of curiosity - why does it launch two
svnserve processes, and is this true for all
repository accesses (i.e. local/file:///, SSH, Apache,
etc.)..?
I will rebuild the server tonight 0.36 and the repos
and try some more tests. Thanks for your feedback.
Regards,
Davis
__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svnserve hang - SMP questions
Posted by Tobias Ringstrom <to...@ringstrom.mine.nu>.
root wrote:
> Hi Tobias,
>
> Thanks for the response. I realize that what you
> indicate below is covered in the FAQ. However, I do
> not believe this to be the case. I am the lone
> administrator of this box and this repository. I can
> set up a brand new clean repository, and then from the
> client, do a checkout and the same thing occurs -- it
> hangs. I can guarantee that there are no other
> clients hitting the repository or local svn* processes
> hitting the repository at the same time the checkout
It's completely OK to have several clients accessing the repository at
the same time of course.
> is underway. Doing svnadmin recover does not seem to
> fix this problem.
I assume that you killed all svnserve:s, ran svnadmin recover, and that
it completed without error. We need to narrow this down a bit further.
Make sure that your repos is not a network drive, and that you do not
have a permission/umask problem.
Can you checkout using file://? If that works, can you run "svnserve -d"
on the server and access it via svn://?
You should also upgrade to the latest release (0.36) of course.
> What I am curious about -- is if it is possibly an SMP
> issue. Has anyone run svnserve on SMP boxes before?
I doubt that this is an SMP problem, but who can know for sure?
> What I see is that a single client doing a checkout
> connects via SSH one time and starts the svnserve
> process, then a second or so later connects via SSH
> again, and launches *another* svnserve process. Is
> this by design -- such that a single checkout requires
> 2 svnserve processes to be running? If so, is it at
Yes.
> all possible that there is deadlock between the two
> processes somewhere along the line that causes this
> hang?
Not likely.
/Tobias
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svnserve hang - SMP questions
Posted by root <eb...@yahoo.com>.
Hi Tobias,
Thanks for the response. I realize that what you
indicate below is covered in the FAQ. However, I do
not believe this to be the case. I am the lone
administrator of this box and this repository. I can
set up a brand new clean repository, and then from the
client, do a checkout and the same thing occurs -- it
hangs. I can guarantee that there are no other
clients hitting the repository or local svn* processes
hitting the repository at the same time the checkout
is underway. Doing svnadmin recover does not seem to
fix this problem.
What I am curious about -- is if it is possibly an SMP
issue. Has anyone run svnserve on SMP boxes before?
What I see is that a single client doing a checkout
connects via SSH one time and starts the svnserve
process, then a second or so later connects via SSH
again, and launches *another* svnserve process. Is
this by design -- such that a single checkout requires
2 svnserve processes to be running? If so, is it at
all possible that there is deadlock between the two
processes somewhere along the line that causes this
hang?
Regards,
Davis
--- Tobias Ringstrom <to...@ringstrom.mine.nu> wrote:
> root wrote:
> > Hi,
> >
> > I have an issue where a checkout of a repos simply
> > hangs.
>
> You need to run svnadmin recover on the repository.
> Make sure that there
> are no other processes running that are accessing
> the repository while
> you run recover.
>
> Problems such as this one occurs if a process dies
> while accessing the
> repository locally (svnserve, svnlook, svnadmin,
> file://, etc). The 0.36
> release has a number of fixes that make these
> processes shut down
> gracefully in cases where they were killed before
> (and thus causing
> repository hangs).
>
> /Tobias
>
__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: svnserve hang
Posted by Tobias Ringstrom <to...@ringstrom.mine.nu>.
root wrote:
> Hi,
>
> I have an issue where a checkout of a repos simply
> hangs.
You need to run svnadmin recover on the repository. Make sure that there
are no other processes running that are accessing the repository while
you run recover.
Problems such as this one occurs if a process dies while accessing the
repository locally (svnserve, svnlook, svnadmin, file://, etc). The 0.36
release has a number of fixes that make these processes shut down
gracefully in cases where they were killed before (and thus causing
repository hangs).
/Tobias
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
svnserve hang
Posted by root <eb...@yahoo.com>.
Hi,
I have an issue where a checkout of a repos simply
hangs.
I am using the following:
TortoiseSVN 0.24.0, Build 554, UNICODE
Subversion 0.35.1, dev build
apr 0.9.5
apr-iconv 0.9.5
apr-utils 0.9.5
berkeley db 4.2.52
neon 0.24
OpenSSL 0.9.7c 30 Sep 2003
zlib 1.2.1
Access is via svn+ssh. Server is Linux
kafka.localdomain 2.4.20-27.9smp #1 SMP
When I do a checkout, things start working fine and
files are copied (these are PDF files). However, at
some point, the network traffic halts, and no further
progress is made. On the server, the svnserve process
is still ready, but not using any CPU. -->
# ps aux | grep svnserve
dford 13534 0.0 0.0 5140 1988 ? S
22:06 0:00 /usr/local/bin/svnserve.bin -t -t
dford 13544 7.8 0.2 7768 4660 ? S
22:06 0:08 /usr/local/bin/svnserve.bin -t -t
# top -d1 -p 13534
PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM
TIME CPU COMMAND
13534 dford 24 0 1988 1988 1708 S 0.0 0.0
0:00 1 svnserve.bin
Same is true for pid 13544 ..both of them seem to be
waiting....forever, for something.
The last two entries in /var/log/messages show:
Jan 15 22:06:24 kafka sshd(pam_unix)[13533]: session
opened for user dford by (uid=500)
Jan 15 22:06:25 kafka sshd(pam_unix)[13543]: session
opened for user dford by (uid=500)
So, I am still logged in and ssh tunnel is still
active. For some reason, tho, the svnserve process
and the client (same for svn cmd line or tortoise) are
not doing anything.
When I kill the client, I can see that I have a zombie
transaction lying around via svnadmin lstxns. I can
clean up all txns using svnadmin rmtxns, but this does
not fix the problem.
On the client side, after killing the checkout, if I
do an update, I will typically make it a bit farther,
and it hangs again. Then I can kill it and do update
again, and it will get a little farther. Eventually,
it will checkout the whole repository, but I am trying
to figure out why it hangs. Unfortunately, I did not
compile this version with --enable-maintainer-mode. I
intend to rebuild the server soon with this option.
I am wondering if there is anywhere else I can look to
try to figure out why this occurs?
Note that I am the only one currently checking in/out
of this repose.
Lastly, I am curious why 2 svnserve processes are
launched when I do a single client checkout? <as
indicated above>
Thank you in advance,
Davis
__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org