Re: svn commit: r14482 - in trunk/subversion: mod_dav_svn

On May 4, 2005, at 4:10 PM, kfogel@collab.net wrote:

> Huh. This seems to be a followup to r14180, but while r14180 made it
> into 1.2.0-rc2, this commit did not (it was merged into 1.2.x after
> rc2 went out).
>
> I'm kind of surprised we're not seeing more complaints about it,
> therefore. Particularly the mod_dav_svn/lock.c:~111 fix -- why
> wouldn't people be getting overflow errors without this change? We're
> more than halfway across the epoch already, after all.
>

Nobody reported it. I merely speculated that it *could* happen.

I was fooling around with autoversioning, and suddenly started getting
weird 500 errors from apache every time mod_dav_svn tried to run
svn_fs_get_locks(). Using gdb, I discovered that *somehow*, a
lock-skel had been written to BDB which contained a creation-date of
January 1914. I have no clue how that happened, and couldn't
reproduce. But that led me to discover that apr_time_t was signed, and
led to r14482.

So really, there's no connection to cmpilato's r14180 change, other
than they're both "commits that have to do with time values". The
potential bug fixed by r14482 existed long before cmpilato's change.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu May 5 01:53:00 2005