Junio C Hamano <gitster@xxxxxxxxx> writes: > Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> writes: > >> On Mon, 4 Aug 2008, Junio C Hamano wrote: >>> >>> I started to revisit this issue and patched "git update-index --add" >>> and "git add" so far. Patches follow. >> >> Patches look good to me, but did you check the performance impact? >> >> The rewritten 'has_symlink_leading_path()' should do ok, but it migth >> still be a huge performance downside to check all the paths for things >> like "git add -u". > > I couldn't quite measure much meaningful performance impact. > > My test repository was the kernel tree at v2.6.27-rc2-20-g685d87f, without > any build products nor editor temporary crufts. > > By the way, if anybody wants to reproduce this, be careful with the tests > that run "rm -f .git/index" before adding everything. After doing so, if > you check the result with "git diff --stat HEAD", you will notice many > missing files --- I almost got a heart attack before inspecting this file: > > $ cat arch/powerpc/.gitignore > include > > Yes, it excludes 261 already tracked files. Is it intended? I dunno. > ... Seeing that you applied my arch/powerpc/.gitignore patch to the kernel (Yaay, I now have a short-log entry in the kernel history ;-), you have seen the message with some benchmarks I am replying to as well? I actually was expecting to see measurable slowdown by checking leading symbolic link more often, but I couldn't, which means either there is not much downside that the bugfix would hurt real-life usage, or I was too incompetent to come up with an example to show that the bugfix actually hurts the performance. If the former, I think it is a very good news (If the latter, well, we have a bigger problem :-<), as the affected codepaths are exactly the ones that need to be fixed when somebody tries to add files in a submodule directory by mistake, which is an issue (iirc) Dscho had a separate patch to fix earlier. Tracked symbolic links may be rare, but it seems that many people are using submodules, and that case cannot be hand-waved around by calling it is a feature. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
