bug-gnulib
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: (sharutils) make distclean removes doc/uudecode.1


From: Bruce Korb
Subject: Re: (sharutils) make distclean removes doc/uudecode.1
Date: Sun, 07 Apr 2013 11:21:42 -0700
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130329 Thunderbird/17.0.5

Hi Bob,

Thank you for your pointers!

Bob wrote:> Bruce Korb wrote:
The last dinkleberry has to do with libintl.
On OS/X:
You have LD_LIBRARY_PATH set including /usr/local/lib?

It was forced (and done by hand) because without it:

> $ unset LD_LIBRARY_PATH
> $ ldd ../src/shar
>         libintl.so.8 => not found

I don't know about OS/X so this response may not be useful.

I confess, I don't develop much on OS/X.  I was granted access
to a mess of build platforms at U of Utah and this is just one.
I ran a script that did the same thing on all platforms:

  mkdir /local/shar-$$ && \
  cd /local/shar-$$ && \
  configure ~/shar-src && \
  make && \
  make check && \
  make install DESTDIR=$HOME/_$(uname -n) && \
  rm -rf /local/shar-$$

To my naive way of thinking, that should just work.
The configure step clearly figures out that libintl
lives in /usr/local, but the build instructions don't
do the right thing to yield a working executable:

It _looks_ like it finds libintl in /usr/local/lib and then
links the executables without marking it with:
   -Wl,-rpath -Wl,/usr/local/lib
Shouldn't that incantation be pretty automatic?

The issue of rpath has been a hot-button of contention over the years.
It is similar to the conflict between searching for executables using
PATH and hard coding the full path to executables.

Rpath should not be needed if the ld.so is configured properly.

OK, so when someone complains I should say, "works as designed"?
Just how hard is it to test to see if /path/to/libintl.so is in
the default link/loader path for a platform?  Then, if not,
either link against libintl.a or do the proper incantation to
make the link/loader look there at run time.  How hard can that
be?  What I _don't_ want is for "configure" to say "all is well"
and then have the program fall over because it really wasn't.

Therefore I would look there for the problem and solution first.

If it were my platform, I'd just fix it.  I am not an admin, but
a guest being given the opportunity to test a release before
releasing it.

I guess I like simple solutions.  "Do thus and so and all is well."

*sigh*.  Thanks - Bruce



reply via email to

[Prev in Thread] Current Thread [Next in Thread]