lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Uwe Schindler" <>
Subject RE: Apache Git mirror
Date Mon, 12 Nov 2012 07:49:25 GMT

> > apply using the standard unix "patch -p1" (to get rid of the crazy a/b path
> prefix, also not existent in the SVN diffs). After that I must take care to not
> forget to add files, adjust svn props because only stock "svn patch" command
> can do this for me.
> Don't want to go into flame wars, but I think it's svn patch that's crippled
> here, not git's. A git patch is pretty standard (?). It also contains file additions/
> removal (via /dev/null referencing), unlike svn's (may be wrong here -- I'm
> not up to date with svn's recent progress). Sure, svn properties are not
> supported but hey -- they're, ehm, svn properties.

>From the point of view of a "patch" in the unix world, both approaches are correct - in
general "a" and "b" looks strange to me and unintuitive. The use of /dev/null or not does
not matter, as UNIX patch is documented to work with both.

My comment was not meant for flamewars, it is just the additional work I have to do and that
it is more error prone and sometimes makes it impossible to apply older GIT patches. The technology
is fine, I just have problems with the usability. If I work on an issue and there is a large
non-SVN patch applied and it is older than 3 days I don't even try anymore to apply the patch.

There is one more useful feature of svn patch for windows users: It uses svn:eol-stlye to
normalize the line endings in the patch with locally used ones. If I apply the patch with
UNIX(Cygwin) patch (without dos2unix or unix2dos), my files suddenly have mixed line endings
and then SVN heavily complains on committing. This drives me crazy... In that case I simply
stop working on the issue and ask the use to resubmit a subversion conform patch (same with
non-unified diffs).

One more thing: If we intend to start testing patches from JIRA with Jenkins (as Grant proposed,
the JIRA plugin for this was implemented by some Hadoop people, as far as I remember), we
need the SVN metadata to cleanly apply patches in heavy-committing times.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message