Re: What's cooking in git.git (Nov 2010, #03; Wed, 24)

[Thomas Rast <trast@student.ethz.ch>]

Jakub Narebski wrote:
> Could you add 'exit' just after second test in 
> t/t9502-gitweb-standalone-parse-output.sh test script, or running it
> with `--immediate' option, and show us the results (after 'cd t') of
> 
>   $ file "trash directory.t9502-gitweb-standalone-parse-output/gitweb.body"
>   should be:
> 
>     trash directory.t9502-gitweb-standalone-parse-output/gitweb.body: tar archive

Well, you're onto something...

  trash directory.t9502-gitweb-standalone-parse-output$ file *
  file_list:          empty
  foo:                ASCII text
  gitweb.body:        empty
  gitweb_config.perl: perl script text executable
  gitweb.headers:     ASCII text, with CRLF line terminators
  gitweb.log:         empty
  gitweb.output:      ASCII text, with CRLF line terminators
  GLOB(0xdf18fc0):    tar archive

Huh.  Seems something got confused about what to use as a filename?

> > OTOH if I check out v1.7.3.1-48-g5768176 instead, where the test
> > works ok, it looks like this: [...]
> 
> Could you bisect to commit that introduces breakage?

That's what the automatic tester does by itself (again under
valgrind), and it came back with "gitweb: File based caching layer
(from git.kernel.org)", i.e., 17b15d4.  I should have pointed out that
this was already a bisection in the first mail.

(The manual runs on the server now verified that the bisecter worked
ok, so it really is that commit.)

-- 
Thomas Rast
trast@{inf,student}.ethz.ch