Re: nasty commit bug

Justin Erenkrantz <jerenkrantz@apache.org> writes:

> Not checking for the return from the hashtable was NULL is incorrect.
> So, r2504 should stay.

I disagree.

The reason it was NULL is that the directory was already locked and
so the attempt to lock it with svn_wc_adm_open failed and nothing got
added to the hash. Your patch just calls svn_wc_adm_open again, which
is going to fail again.

The point is it should *never* be NULL. If you want to check it use an
assert or return an svn_error_t. Continuing under these circumstances
is wrong. You have added obscure code that is extremely unlikely to
ever work. If by chance it does work then it is hiding another bug.

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org