[cvsnt] Re: Explanation of crash on cvsnt server?

[Daniel Rabe]

Hey, this sounds like what I'm seeing! I'm running a script to try to
convert our SourceSafe database to CVS. It's crashed this way about three
times in the last week or so. I tried sending the crash dump to the
crash-dumps email address, but it's too large. (The crash dump is 14Mb, and
the mailbox apparently has a 10Mb limit.)

--Dan

"Christian Ottosen" <cso at mita-teknik.com> wrote in message
news:b3iarh$abf$1 at sisko.nodomain.org...
> REQUEST:
> - Have anyone a good explanation of this result (see RESULT)?
> - Is it a realistic test setup (see SETUP)?
> - why does many commits in a row result in a cvs server crash?
>
> RESULT:
> After modifying and committing the same text file over and over again (in
a
> script) the cvsnt server crashes. The crash happens after 10 - 100 commits
> (varies for each test). The cvsnt server crashes with this error message:
>
> "Something bad happened to CVSNT, and it crashed. Would you like to
produce
> a crash dump?"
>
>
> SETUP:
> In order to test if the CVSNT server is stable we have done this:
>
>   1. Installed cvsnt (1.11.1.3 beta build 72) on W2k server
>   2. Installed WinCvs (1.3b12) on W2k client
>   3. Imported a test module
>   4. made a tcl script (se below), which edits and commits a textfile over
> and over again
>   5. finally we run this script through WinCvs using tcl shell command
> 'source script'
>
> Content of 'script':
> for {set x 0} {$x<500} {incr x} {
>     exec cmd.exe /c echo 1 >> file2.txt
>     cvs -x commit -m "no message" file2.txt
> }
>
>