Re: make ChangeLog dep.h filedef.h read.c remake.c ...

[Paul Smith]

On Wed, 2009-09-30 at 19:51 +0200, Boris Kolpackov wrote:
> All four bugs were variants of the "dependency failed to update while
> rebuilding -include'ed makefile while another, real, target depended
> on it but make didn't issue any or issued bogus diagnostics". I only
> could think of three tests for this.

Sounds good.

> > So, the question becomes: do you think we're ready for rc1 tomorrow?  Or
> > are there more things to do such that we should hold off?
> 
> I would really like to do some profiling. Plus take a look at #25780 and 
> #25140. So maybe wait another week?

OK, that's fine.  I am uber-busy this week anyway.

FYI, I added some regression tests for #25780 a few days ago; if you run
the test suite with -all you'll see them failing [*].


-----
[*] I added a feature to the test suite where we can define tests inside
"if ($all_tests) ..." which won't be run normally, but only be run if
-all is given to run_make_tests (or you can run "make check
MAKETESTFLAGS=-all").

What I'd like to do is get to triaging make bugs more quickly, where
when a bug comes in we check it out and see if we can reproduce it
fairly soon after the bug is reported.  If not we can ask for more info.
Once we understand the bug we can set its "Triage Status" to Verified
or, if we have some thought about how hard it is, to one of the other
values.  And, if possible, we can add a regression test to reproduce it
to the test suite, delimited by the $all_tests check so it doesn't cause
the tests to fail.

We'll see how it goes :-).