bug#25317: emacs-25.1.91 build feedback

bug#25317: emacs-25.1.91 build feedback

[Nelson H. F. Beebe]

I'm pleased to report that in our test lab of 130+ Unix-(like) and
other O/Ses, I have now done 175 build attempts of emacs-25.1.91, and
of those, 109 have succeeded.

I'm gradually working my way through the failing builds, trying
alternative compilers, and configure options.

However, 2016 has drawn to a close, and it is time to go home to
celebrate New Year's Eve.

I'll report more early next week on the platform-specific build issues
that I've identified.


-------------------------------------------------------------------------------
- Nelson H. F. Beebe                    Tel: +1 801 581 5254                  -
- University of Utah                    FAX: +1 801 581 4148                  -
- Department of Mathematics, 110 LCB    Internet e-mail: beebe@math.utah.edu  -
- 155 S 1400 E RM 233                       beebe@acm.org  beebe@computer.org -
- Salt Lake City, UT 84112-0090, USA    URL: http://www.math.utah.edu/~beebe/ -
-------------------------------------------------------------------------------

bug#25317: emacs-25.1.91 build feedback

[npostavs]

"Nelson H. F. Beebe" <beebe@math.utah.edu> writes:
> I'll report more early next week on the platform-specific build issues
> that I've identified.

ping?

bug#25317: emacs-25.1.91 build feedback

[Nelson H. F. Beebe]

>> ping?

I've been extremely busy the past 2 weeks with start of term, and
today, a power outage and snow storm that has kept me busy all day
restoring machines.

I have succeeded in getting several of the emacs-25.1.91 failures to
build, by changing compilers / options / PATH.  There remain a few
that are as yet unresolved.  I'll try to get to those next week.

One pain in emacs-x.y.z releases is that if any of the graphics
libraries (gif, jpeg, png, tiff, ...) and associated header files are
missing, configure aborts.  

I would MUCH rather that it just warned that it would not be able to
supply that rarely-used feature of viewing bitmap graphics files, and
carry on with the build.  

I had to manually restart a substantial fraction of the builds that
I've done so far (157 automated, 53 manual).  On several platforms, it
seems difficult to get all of the needed graphics libraries installed,
without running into package dependencies that say "you can have THIS
package, but not THAT package, because it depends on YET ANOTHER
package that cannot be installed on your O/S version".

-------------------------------------------------------------------------------
- Nelson H. F. Beebe                    Tel: +1 801 581 5254                  -
- University of Utah                    FAX: +1 801 581 4148                  -
- Department of Mathematics, 110 LCB    Internet e-mail: beebe@math.utah.edu  -
- 155 S 1400 E RM 233                       beebe@acm.org  beebe@computer.org -
- Salt Lake City, UT 84112-0090, USA    URL: http://www.math.utah.edu/~beebe/ -
-------------------------------------------------------------------------------

bug#25317: emacs-25.1.91 build feedback

[npostavs]

"Nelson H. F. Beebe" <beebe@math.utah.edu> writes:
>
> One pain in emacs-x.y.z releases is that if any of the graphics
> libraries (gif, jpeg, png, tiff, ...) and associated header files are
> missing, configure aborts.  
>
> I would MUCH rather that it just warned that it would not be able to
> supply that rarely-used feature of viewing bitmap graphics files, and
> carry on with the build.  

Hmm, the problem is that configure.ac is currently structured in such a
way that it can't distinguish between --with-gif=yes and not passing any
--with-gif option, so changing this is not so easy.  As a quick fix, you
can apply the following patch to disable this checking for all the image
libraries:

--- i/configure.ac
+++ w/configure.ac
@@ -3517,7 +3517,7 @@ fi
 AC_SUBST(LIBGIF)
 
 dnl Check for required libraries.
-if test "${HAVE_X11}" = "yes"; then
+if false ; then
   MISSING=""
   WITH_NO=""
   test "${with_xpm}" != "no" && test "${HAVE_XPM}" != "yes" &&

Or if you're building from the tarball without running autoreconf, patch
configure directly:

--- i/configure
+++ w/configure
@@ -15731,7 +15731,7 @@
 fi
 
 
-if test "${HAVE_X11}" = "yes"; then
+if false ; then
   MISSING=""
   WITH_NO=""
   test "${with_xpm}" != "no" && test "${HAVE_XPM}" != "yes" &&

bug#25317: emacs-25.1.91 build feedback

[Eli Zaretskii]

> Date: Sat, 21 Jan 2017 18:11:43 -0700
> From: "Nelson H. F. Beebe" <beebe@math.utah.edu>
> Cc: 25317@debbugs.gnu.org, beebe@math.utah.edu
> 
> One pain in emacs-x.y.z releases is that if any of the graphics
> libraries (gif, jpeg, png, tiff, ...) and associated header files are
> missing, configure aborts.  
> 
> I would MUCH rather that it just warned that it would not be able to
> supply that rarely-used feature of viewing bitmap graphics files, and
> carry on with the build.  
> 
> I had to manually restart a substantial fraction of the builds that
> I've done so far (157 automated, 53 manual).  On several platforms, it
> seems difficult to get all of the needed graphics libraries installed,
> without running into package dependencies that say "you can have THIS
> package, but not THAT package, because it depends on YET ANOTHER
> package that cannot be installed on your O/S version".

The behavior was originally like you prefer, but it was changed to
what it is now because users complained that a warning amidst the gobs
of output the configure script pours on users is very hard to notice.

I see your point, but your situation is an unusual one, whereby you
build many different configurations via automated scripts.  That's not
how most users build Emacs.

Perhaps a special option --with-PACKAGE=silent could be added to cater
to your use case.  But it cannot be the default, IMO.