Re: outdated m4sugar

[Eric Blake]

Joel E. Denny <jdenny <at> ces.clemson.edu> writes:

> > Again, rather than creating a new file, I'd copy what autoconf does (it
> 
> > Or all three.  Would you like me to propose a patch?
> 
> Sure!

I'm waiting for my copyright paperwork to complete, at this point.  But on a 
whim, I did 'cp autoconf/lib/m4sugar/m4sugar.m4 bison/data/m4sugar', and reran 
the testsuite, to see where the problem spots would be in a wholescale upgrade 
to the latest m4sugar, and only got the following new failures:

  49: output.at:188      Output file name: address@hidden&*()-=_+{}[]|\:
      <>, .'
  50: output.at:194      Output file name: (
      c++
  51: output.at:195      Output file name: )
      c++
  52: output.at:196      Output file name: #
      c++
  53: output.at:197      Output file name: @@
      c++
  54: output.at:198      Output file name: @{
      c++
  55: output.at:199      Output file name: @}
      c++
  56: output.at:200      Output file name: [
      c++
  57: output.at:201      Output file name: ]
      c++
 191: c++.at:159         Relative namespace references
      c++
 192: c++.at:165         Absolute namespace references
      c++

Plus 11 skips because I didn't have doxygen or a java compiler installed on the 
machine where I was testing.

49-57 are probably all related to a subtle change in the quoting semantics of 
at least one of the (possibly-undocumented) m4sugar macros, such that the 
filename is now being expanded where it used to be handled as a quoted string.  
I'm not sure about 191 or 192, but hope it is the same failure, as a more 
detailed scan of testsuite.log showed that the preprocessor error messages 
looked similar on the attempt to compile the bison output.  At any rate, it 
means that for now, we are better off doing a partial sync from autoconf's 
m4sugar, rather than a whole-scale upgrade, until I can isolate which piece(s) 
of the upgrade caused the additional test failures.

Thankfully, 'git add -i' makes it easy to pick and choose which pieces of 
m4sugar.m4 I want to merge at this time, before posting my patch.

> > Good comment.  But now you need to do 'gnulib-tool --import unsetenv'.
> 
> Ok, I don't mess with gnulib much.  If I put unsetenv in gnulib_modules in 
> bootstrap.conf, then lib/unsetenv.c is created but not lib/unsetenv.o.  
> I'm out of time.  Does this mean unsetenv isn't needed on my platform, or 
> does this mean I'm not doing this right?

You did it right.  It means that your platform's unsetenv is not broken, so the 
replacement unsetenv.c did not have to be compiled (looking at config.log or 
config.h can confirm this fact).

-- 
Eric Blake