Re: svn commit: r1202783 - /subversion/trunk/subversion/libsvn_fs_fs/fs_fs.c

Daniel Shahaf <d.s_at_daniel.shahaf.name> writes:

>> fs_fs knows that in this case the root cause is important and that the
>> error chain is mostly useless. mod_dav_svn, and svnserve, cannot know
>> that so would have to print the whole chain in every case.
>
> "In this case"? How do you know what the case is? It might be a disk
> problem, not just the specific error that triggered your fix.

Getting the rep from the rep-cache database failed. The root error is
the one that matters, whether it is "permission denied" or "wrong sqlite
version" or "i/o error". Recording "failed to open database" is not
useful.

> Should we add some SQLite error code to the if() condition whose else()
> branch you patched? For example, "SQLite db is corrupted" would appear
> to be serious enough (if it's not already caught by the
> SVN_ERR_FS_CORRUPT check).

That branch doesn't clear the error, I assume it gets handled elsewhere.

-- 
uberSVN: Apache Subversion Made Easy
http://www.uberSVN.com