Re: pr19.test

[Tom Tromey]

>>>>> "edward" == edward  <address@hidden> writes:

I'm finally reading this thread.
I haven't finished it yet, so what I say may make little sense :-(

edward>     # FIXME: nodist.
edward>     &push_dist_common ($pfx . $base . '.' . $ext);

edward>     commenting out the last line removes foo.c (which is a
edward> temporary file) from the make distdir target, which fails
edward> because according to automake, foo.c is supposed to be
edward> redistributed. however, (gnu) make removes temporary files
edward> like foo.c (generated from foo.l). in my opinion, this is
edward> correct and should not be redistributed. the Makefile.am
edward> writer can always explicitly include foo.c if they really want
edward> to. from the comment in the automake.in file, it seems the
edward> automake peeps are aware of this. there are similar comments
edward> in automake.in with other generated files as well.

Some notes:

* The GNU coding standards require us to distribute these files.
  Since the primary goal of automake is to implement these standards
  (as much as possible), we have to do this.

* The `FIXME: nodist' comment refers to a case like this:

    bin_PROGRAMS = foo
    nodist_foo_SOURCES = parse.y

  In this case neither parse.y nor parse.c should be distributed.
  Currently, I think, we fail this.

edward> Here is the context:
edward> foo.o: foo.l
edward> make has enough information to convert foo.l -> foo.c (via
edward> lex/flex) then compile foo.c to foo.o. However, since foo.c is
edward> an intermediate file it is deleted.

I'm confused on this point.  Why don't we see this failure on
platforms other than Cygwin?

Tom