Robert Schwebel <[EMAIL PROTECTED]> writes:

> first of all, the program_invocation_name problem was indeed a bug in
> uClibc. glibc sucks the variable in from libc.map, whereas uClibc doesn't
> have support for program_invocation_name. The declaration in errno.h came
> from the original glibc errno.h. In the current CVS this was removed by
> the maintainer.

So that problem is solved? Good. I still don't quite understand what
was going on. Did the old errno.h miss an "extern" in the declaration
of program_invocation_name? I.e.

  char *program_invocation_name; /* A "tentative definition" */

instead of

  extern char *program_invocation_name; /* A declaration */

> > The configure script should figure out whether or not you have it, and
> > adapt the build to the situation. I still want to see what you get in
> > your config.log (and if rerunning configure, you may have to delete an
> > config.cache first, or the tests will be omitted).
> 
> I'll test this again tomorrow, because I have to downgrade uClibc to the
> version from yesterday.

It looks like the testprogram was compiled and linked with no warnings
and no errors.

> Attached you'll find my patch which lets you compile lsh-1.2.5 with both,
> glibc and uClibc.

Thanks! I'll add those changes in both the 1.2.x and 1.3 branches.

> With glibc there is still the problem with addprefix, but if I
> remember correctly you had already found a solution for that.

No, I haven't solved that. If I recall the problem correctly, the
situation was that your glibc includes a good enough version of argp,
so that LIBARGP is empty (instead of "argp/libargp.a", like it is on
my systems).

Then, in src/testsuite.Makefile.am, $(addprefix ../,@LIBARGP@) is
expanded to something bogus, like "../".

What is LIBARGP? Does it matter if you replace @LIBARGP@ with
$(LIBARGP)? (There's a line LIBARGP = @LIBARGP@ earlier in the
generated Makefile.in, so it shouldn't really matter).

>From reading the make manual, and some testing with the make I have
installed (GNU make 3.79.1), it seems that $(addprefix ../,) expands
to the empty string. I guess that's why I used addprefix from the
start, instead of ../@LIBARGP@. So I don't see why it goes wrong. What
version of make are you using?

Regards,
/Niels

Reply via email to