[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: errors on large commits

From: John Waycott <javajohn_at_cox.net>
Date: 2006-09-16 16:07:45 CEST

Garrett Rooney wrote:
> On 9/14/06, Rob Brandt <bronto@csd-bes.net> wrote:
>> I am having exactly the same issue as Markus Fischer a few months ago
>> where the error message is:
>>
>> svn: Commit failed (details follow):
>> svn: MERGE request failed on '/repos/web/project/trunk'
>> svn:
>> Cannot modify transaction named '1hv' in filesystem
>> '/data/subversion/repos/db'
>>
>> See:
>> http://subversion.tigris.org/servlets/ReadMsg?listName=users&msgNo=47260
>>
>> However his solution is not helping me. I tried both the --charset on
>> SVN::Notify and using a fresh checkout. The repository is exhibiting
>> the same symptoms; the commit is actually successful at the server end,
>> but my own working copy doesn't know that. My workaround is to revert
>> after the commit, delete the new files (at the moment I'm just trying to
>> load source files into a new repository) and update. That's not a long
>> term solution :(
>>
>> My own personal observation is that the commit only fails on large
>> commits. Not once has it failed when the commit is less than 100k. So
>> far, I haven't seen success above 100k, but most of the one above 100k
>> are much larger than 100k, so I'm not sure where the dividing line is.
>> This does appear to be an SVN::Notify issue, as I disabled that in the
>> post_commit hook and it doesn't choke.
>>
>> Ideas?
>
> Do you happen to be running mod_php on the same server?
>
> -garrett
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: users-help@subversion.tigris.org
>
>
One of our users got this message about two weeks ago. I ran svnadmin
verify and found a corrupted revision. This doesn't make sense, but
fsfsverify fixed the problem. I don't see how that error could be caused
by that, but the user could commit afterwards. She was committing a very
large 7MB file. Maybe this was just coincidence?

We've hit the fsfs corruption bug five times in the last two months, so
it is fairly frequent. The error indication in other cases was the "read
length line" error.

-- John

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Sat Sep 16 16:08:21 2006

This is an archived mail posted to the Subversion Users mailing list.