Re: GNU make release candidate 4.1.90 available for download

[Eli Zaretskii]

> Date: Sun, 24 Apr 2016 10:01:26 +0300
> From: Eli Zaretskii <address@hidden>
> Cc: address@hidden
> 
> > From: Paul Smith <address@hidden>
> > Date: Sat, 23 Apr 2016 17:32:15 -0400
> > Cc: address@hidden
> > 
> > A new release candidate for GNU make 4.2 is available now for download:
> > 
> >     90a3ac4fbf4a1cf8dacaf92b079b16b2  make-4.1.90.tar.bz2
> >     ffd56d762220269c0534edfda9197cb3  make-4.1.90.tar.gz
> > 
> > You can obtain a copy from:  http://alpha.gnu.org/gnu/make/
> 
> Thanks.  It didn't compile with MinGW GCC 4.9.3, due to use of C99
> features.  I pushed a fix.  I also fixed a potential bug in printing
> time_t values using %d format descriptor, which would be wrong when
> time_t is a 64-bit type; this was uncovered by a compiler warning.
> 
> I will run the test suite later.

Did that now, using MSYS Bash and MSYS Perl.  Bottom line: all the
features that are supported on MS-Windows pass the tests.

I needed a few changes in the test suite, to get it to reports as few
bogus failures as possible, as expected.  Here are the most prominent
issues:

 . The 'load' and 'loadapi' tests hard-code the .so extension of the
   loadable modules, and don't have a provision for using the import
   library in the link command, as required on Windows.

 . The 'abspath' and 'relpath' use Posix-style absolute file names
   that begin with a slash, and expect results that begin with a
   slash, which doesn't happen on Windows.

 . Tests that use Posix shell features too literally (like $$var)
   fail.  Could this be tested using the 'env' command instead?

 . Some tests check for messages too literally.  Example:

     *** work/features/archives.base    Sun Apr 24 15:39:26 2016
     --- work/features/archives.log     Sun Apr 24 15:39:26 2016
     ***************
     *** 1,3 ****
       ar Urv libxx.a a1.o
       a - a1.o
     ! d:\usr\bin\ar.exe: creating libxx.a
     --- 1,3 ----
       ar Urv libxx.a a1.o
       a - a1.o
     ! ar: creating libxx.a

 . Last, but not least, it would be nice if the test suite would by
   default test the Make binary just built, instead of requiring me to
   use -make and -srcdir options.  (Not sure if this one is Windows
   specific.)

Thanks.

P.S. I would like to thank Paul for keeping the Windows build scripts
up to date with the changes in the source structure.  Awesome work!