This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Patch: automatic dependencies for gcc


* Tom Tromey wrote on Tue, Mar 25, 2008 at 08:38:37PM CET:
> >>>>> "Ralf" == Ralf Wildenhues <Ralf.Wildenhues@gmx.de> writes:
> 
> Ralf> I forgot one more thing here.  If gcc is interrupted, it may leave an
> Ralf> incomplete dependency file here, which can lead to make erroring out
> Ralf> next time (make won't remove the .Po file upon interrupt, as it's not
> Ralf> listed as target).  You could use $(*D)/$(DEPDIR)/$(*F).Tpo and a
> Ralf> subsequent mv to *.Po as automake does.
> 
> Thanks.  I'll look into this.

As a really minor note, if you can make it two separate command lines
(instead of '$(COMPILE) ... && mv ...') you'll be able to avoid a
subshell (Automake 1.10 does this).

> I suspect we ought to fix gcc as well, though I'm concerned about
> compatibility.  It seems weird to have a bunch of -M* options
> specifically added for this automake feature, and then have to work
> around an oddity.  Perhaps we could add yet another -M option to
> finesse the compatibility problem... and I guess add another special
> case to automake.  I dunno.

Why is another -M* switch needed?  IOW: why would atomic creation of the
file listed by -MF be a backward incompatibility, except maybe for the
issue that a fixed gcc is difficult to detect with a feature test?

> Ralf> Yet another thing: you removed this from objcp/Make-lang.in:
> 
> Thanks.  I will fix this shortly.
> I'm doing yet another build to try to reproduce it first.

Good thinking.  So far I only got it on an incremental build, the full
rebuild is still running.

Thanks,
Ralf


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]