innbind fatal error

Russ Allbery rra at stanford.edu
Thu May 8 17:27:53 UTC 2008


Julien ÉLIE <julien at trigofacile.com> writes:

> I now understand what is going on.
>
> In order to reproduce the issue:
>
>     ./configure --enable-ipv6
>     make
>     make check
>     ./configure
>     make
>     make check

Ah, yes, that explodes.

> I think it is because "make depend" is missing in the test suite but
> it does not seem easy to add in test/Makefile...

Yeah, exactly.  I thought about trying to fix that and gave up.  :)

Since I redid the INN build system, I've become more of a fan of using
Automake in its non-recursive mode.  It might be worth switching INN over
at some point since it would fix problems like that, although it would be
a fair bit of work.

> lib/network.c only emits a warning and backends/innbind.c a fatal
> message to LOG_CRIT.

I could have sworn that at some point I'd put some code into innbind so
that it wouldn't log to syslog if an environment variable indicating that
we're testing was set, but I'm apparently imagining things.

> Maybe message_handlers_die() should be added to libtest.c (which only
> captures notice and warn).  But it does not solve the issue of
> dependencies.  So maybe "make check" should run "$(MAKE) clean &&
> $(MAKE) build && ./runtests TESTS", which would solve the whole problem
> (it does not affect only innbind).

That would work and would be the simplest solution.

Another solution would be to change the way the test suite is built to
match what I've started doing elsewhere.  Right now, the test suite builds
another copy of a lot of files in lib with TESTING set and does so with
special Makefile rules.  When I moved to Automake on other projects, I
found that was painfully hard to do, so instead I started creating
wrappers in the test suite directory that look like:

==> asprintf.c <==
#define TESTING 1
#include <portable/asprintf.c>

If you do this, then gcc -MM will get the dependencies right and you can
probably put real dependencies into the tests/Makefile.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.


More information about the inn-workers mailing list