Re: Bug#309185: tetex-bin: FTBFS in experimental: Endless loop.

[Florent Rougon <f.rougon@free.fr>]

Hi,

Hilmar Preusse <hille42@web.de> wrote:

> Something similar happens to tetex-base, when trying to build on 2.6.
> I'm cloning and reassigning to tetex-base. The above analysis tells
> me, that the build will fail on 2.4 too, if the first and the second
> step will happen in different seconds (by chance). Is that correct?

Hmm. It builds[1] on 2.4, though debian/rules is regenerated twice
before the build really starts. The mechanism is slightly different than
in tetex-bin, and, I must say, clever[2]. The .PHONY targets are really
phony targets; the logic looks 100% correct to me.

I'll have to think more about that because I cannot see at the moment an
explanation for debian/rules being regenerated twice that doesn't
conflict with the hypothesis I used to explain why the loop breaks on
2.4 kernels for tetex-bin (i.e. : whether Make will rebuild a target
that has exactly the same timestamp as the most recent of its
prerequesites)...

Could you please insert the following lines:

	@printf "\n<Debug>\n"
	@ls -l --time-style=full-iso $(eperl_sourcefiles) \
          debian/rules.in debian/rules
	@printf "</Debug>\n\n"

after:

# Eperl is simply great: thanks, Davide!
% :: %.in commonstuff_update $(eperl_sourcefiles) 

in debian/rules.in and debian/rules, then run:

  fakeroot debian/rules clean >/tmp/log 2>&1

and post the contents of /tmp/log so that we can better understand what
happens with those timestamps?

Thanks.


[1] Well, "fakeroot debian/rules clean" works. I didn't try a full
    build.

[2] I think "touch -r" could be used instead of the longer
    "touch -t `ls -l --time-style=+%Y%m%d%H%M.%S etc.", but the
    priority for the moment is to fully understand what happens and fix
    it.

-- 
Florent