Re: What's cooking in git.git (Nov 2010, #02; Wed, 17)

Jakub Narebski [off-list ref] writes:

Sidenote: recently sent

  gitweb: selectable configurations that change with each request

practically reverts

  gitweb: Move call to evaluate_git_version after evaluate_gitweb_config

Just FYI.

Hmph, will have to look at it again.

quoted

* jn/gitweb-time-hires-comes-with-5.8 (2010-11-09) 1 commit
 - gitweb: Time::HiRes is in core for Perl 5.8

Looked reasonable.  Will merge to next.

Thanks. With or without improvement to commit message?

I think what I pushed out has already been reworded.  Please check.

quoted

* jh/gitweb-caching (2010-11-01) 4 commits
 . gitweb: Minimal testing of gitweb caching
 . gitweb: File based caching layer (from git.kernel.org)
 . gitweb: add output buffering and associated functions
 . gitweb: Prepare for splitting gitweb

Temporarily ejected while I shuffled jn/gitweb-testing; will queue the
latest back in pu or perhaps in next.

The advantage of 'gitweb: File based caching layer (from git.kernel.org)'
is that it is tested in real-life on heavy load (assuming that 
git.kernel.org uses the same version as is/would be in pu/next).

The disadvantage is that it is seriously messy code.  Something that I
wanted to improve in my rewrite.  This is only minimal fixup.

Which is exactly what we want at this point (I want to release 1.7.4 by
the end-of-year holidays, which means a feature-freeze will have to start
soon).  My understanding is that the serious messiness does not come from
the caching layer.

I am thinking about splitting main 'gitweb: File based caching layer
(from git.kernel.org)' patch in two, separating moving test for
$caching_enabled out of cache_fetch to separate commit (largest change
to original J.H. submission), but leaving hardening "do 'cache.pl';"
and replacing 0/1 valued $cache_enable with boolean valued 
$caching_enabled.

Because currently new tests in t9501 and t9502 (examining status and
output of gitweb with caching enabled) do not pass, I am thinking
about adding new configuration know turning off "Generating..." page.

BTW. should I forge J.H. signoffs, and add mine?

Just ping him beforehand ;-)