Menu
▾
▴
4 Attachments
#743 Patch series to add new configure options and fix issues
This patch series adds some new options to the configure script and also adresses some issues:
- Add configure options to enable/disable docs.
- Add configure options to enable/disable demos.
- Remove LISPDIR from SUBDIR since it doesn't exist anymore.
- Fix usage of gdlib. Use the one that configure script has found.
The gdlib-config patch breaks the build system for me.
./configure fails with messages:
./configure: line 10755: GDLIB_CONFIG: command not found
./configure: line 10756: GDLIB_CONFIG: command not found
In config.log I see
GDLIB_CONFIG='/usr/bin/gdlib-config'
and
ac_cv_path_GDLIB_CONFIG=/usr/bin/gdlib-config
But this definition is apparently not passed on correctly.
This has the effect of clobbering LDFLAGS in the make files that are generated, and at that point I can't recover a working build without reverting the patch and starting over.
Those errors are cause by silly missing '$' characters where the variable is referenced.
0001 disable docs
0002 disable demos
With these patches applied I get the following WARNINGS from ./configure:
=== configuring in docs (/home/merritt/cvs/gnuplot-problem/docs)
configure: WARNING: no configuration information is in docs
=== configuring in demo (/home/merritt/cvs/gnuplot-problem/demo)
configure: WARNING: no configuration information is in demo
Those are caused by the AC_CONFIG_SUBDIR invocations added by those patches. Which are entirely superfluous. gnuplot does not run 'configure' in any subdirectories any more.
I would object to both patches for a different reason, though. Building a program, but not the documentation, is just fundamentally wrong. And adding special configury just to turn off the building of one more tiny program, and its execution, seems excessive.
Last edit: Hans-Bernhard Broeker 2016-12-01
Patches 0001 and 0002 updated to not use AC_CONFIG_SUBDIR.
I don't think that building a program and not building its documentation is fundamentally wrong. You may want to save disk space by not including the documentation in your target device (I'm thinking about embedded devices), so in that case building the documentation is a waste of effort. Why not give the user the possibility to decide?
That can be reason not to install the documentation. It's way insufficient as a reason to not even build it.
I agree.
And if you were to disable the demos you would lose the target "make check" which confirms a successful build by running demos as unit tests. So that's a bad idea also.
0003 remove LISPDIR
No problem with this one
Patch 0004 updated. Added missing "$" in variable name.
3 + 4 applied
1 + 2 rejected