Re: flex beta 2.5.22 released

[W. L. Estes]

On Friday, 11 October 2002,15:13 +0200, Akim Demaille wrote:

> I don't get it: your suite runs properly, so what is the problem with
> my patch?

Ultimately, all the programs in the test suite should be
check_PROGRAMS, and I should the automake testsuite facilities, right?
(I believe the automake manual refers to this is a "simple test suite".)

> The point is probably not usefulness: I was not saying that yours was
> not.  But more of readability.  It is simply that reading ChangeLog is
> so frequent, that it's better to have properly formatted etc.  rcs2log
> does not help making it nice to the reader.

I'm right with you on this one. I'd rather the ChangeLog be more
usefule as opposed to less useful. flex has one now so it's all ok. If
I can make it easier for people within reason, I will.

> Another point is: wrapping a patch with a ChangeLog is easy.  Writing
> a patch with a RCS log is not :)

Yeah, that has bothered me... Can you provide urls for the cvs-utils
or cvsutils package? (whichever is relevant.)

> | What I'm doing with dist_subdirs is this: I don't want "make all" to
> | recurse into the test suite. Am I doing this wrong?
> 
> It's customary to being able to run make check from the top level, so
> Automake has to recurse into your tests directory.

Ok, noted. Thanks.

> | -noinst_SCRIPTS = \
> | +dist_noinst_SCRIPTS = \
> |     create-test
> | 
> | It's not at all clear that noinst_SCRIPTS will not be included in the
> | distribution. (I just read the section in the automake 1.7 manual.)
> 
> The documentation should certainly be adjusted!

Thanks.

> | Also, the tests/TEMPLATE directory does not need to be run when making
> | test/check/all, etc. Am I approaching that problem in the wrong way?
> | 
> 
> Sorry, I don't know what you mean here.

tests/TEMPLATE/ contains files for making tests for the test suite. It
does not contain a test which should be run. So I'd like to have
automake avoid doing anything in there except for dist and distdir,
etc. Am I doing this the right way?

> I'm acheiving that make check fails when the test suite fails.

Is that what automake expects? In my way of thinking, if the test
suite ran but some of the tests failed, that's not a make failure. (If
this is covered in the automake manual, just point to me to the
section I should read.)

> | -BISON = @BISON@
> | 
> | If we dont have this, automake wants to have bison be "bison -y" we
> | want bison with no options or anything. How can we get this?
> 
> You don't have to add these lines: Automake knows its jobs, and knows
> that if there is an AC_SUBSTS(BISON), then it needs to insert these
> lines.

Ok, I'll try, but I remember having bad luck with this.

> |  FLEX = $(top_builddir)/flex
> | 
> | -builddir = @builddir@
> | 
> | If I don't define this, $(builddir) shows up empty, despite what the
> | automake/autoconf manual says. Is that a bug in one of those two packages?
> 
> Well, in that case, you do have found a bug!  But I can't see it:
> 
> src/flex-2.5.22/tests % fgrep 'builddir =' **/Makefile.am        nostromo 
> 15:12
> src/flex-2.5.22/tests % fgrep 'builddir =' **/Makefile           nostromo Err 
> 1
> Makefile:top_builddir = ..

that's top_builddir, not builddir. I need builddir to be defined as
well as top_builddir.

--Will