Re: svn commit: r1935862 - in subversion/branches/svnsync-atomic-author/subversion: include include/private libsvn_ra_local libsvn_ra_serf libsvn_ra_svn libsvn_repos mod_dav_svn svnserve svnsync tests/cmdline tests/libsvn_repos

On 4. 7. 2026 12:59, Evgeny Kotkov via dev wrote:
> Jordan Peck<[email protected]> writes:
>
>>    Read 1 file
>>
>> ● Update(subversion\libsvn_repos\fs-wrap.c)
>> Added 1 line, removed 1 line
>>                                     svn_repos_t *repos,
>>                                     svn_revnum_t rev,
>> _t *pool)
>> {
>>    apr_array_header_t *revprops;
>>
>> ● Update(subversion\libsvn_repos\fs-wrap.c)
>> Added 1 line, removed 1 line
> …
>> Fix 5c — state commit_callback nullability on editor6 directly
>> (replacing the stale pre-1.6 parenthetical):
>>
>> ● Update(subversion\include\svn_repos.h)
>> Added 1 line, removed 2 lines
> …
>> You've hit your session limit · resets 9:50pm (Europe/London)
>> /upgrade or /usage-credits to finish what you’re working on.
>>
>> ✻ Baked for 5m 21s
>>
>> ❯ Can you put together a brief commit message for this change
>>
>> Here's a commit message in the Apache Subversion log-message convention
>> (summary paragraph plus per-file annotations), ready for svn commit -F:
> Looks like the commit message accidentally picked up part of your Claude
> session, not just the generated message itself.


I'm going to be less diplomatic. We know how LLMs are prone to 
"hallucination". And yet you didn't bother to review the commit message 
it prepared for you. How do we know that the code wasn't generated with 
a similar attention to detail?

This is open source. You're never in too much of a hurry to get things 
committed. We prefer ... or rather, I prefer quality over quantity. 
Please do better next time.

-- Brane