Re: New bugs

[Akim Demaille]

>>>>> "Tom" == Tom Tromey <address@hidden> writes:

Tom> Akim -- Your recent patches broke a couple of test cases.

Tom> I looked at the `subobj4' failure.  I think the new code in
Tom> add_depend2 is wrong.

Tom> First, this is wrong:

Tom>             . &transform ('$(' . $pfx . 'COMPILE)' => $rule,

I looked at depend2.am:

@address@hidden:
@AMDEP@ source='@SOURCE@' object='@OBJ@' libtool=no @AMDEPBACKSLASH@
@AMDEP@ depfile='$(DEPDIR)/@address@hidden' 
tmpdepfile='$(DEPDIR)/@address@hidden' @AMDEPBACKSLASH@
@AMDEP@ $(@address@hidden) $(depcomp) @AMDEPBACKSLASH@
        $(@address@hidden) -c -o @OBJ@ `test -f @SOURCE@ || echo 
'$(srcdir)/'address@hidden@

And I'd like to ask a question: why not using @COMPILE@ instead of
replacing $(@address@hidden)?  It seems cleaner to me, and less
surprising for Sunday hackers :)


Tom> Later we see:

Tom>             # Generate rule for `.o'.  . 's/address@hidden@\.o:/' . $obj
Tom> . '.o: ' . $source . '/g;'

Tom> I think we need to quote $obj and $source here; this was handled
Tom> in the old code.

I'll handle this at the same time, when I know what kind of changes
I'm allowed to.  BTW, I'd like to commit the file_contents patch first
if you OK it, as it will diminish the workload.


Tom> I didn't investigate the specflags5.test failure.  Could you do
Tom> that?

I sure will.  In fact, I did see those failures, but didn't realize
they were mine :(.

Tom, would like me to install something like what I did for Autoconf:
betas display a banner extracted from BUGS, stating if can be used, or
must not.