- Use EL_ARG_DEPENDS(), the other way ended up in undefined symbols
errors with --enable-fsp and no fsplib.
- Also check for -lfsp in addition to -lfsplib.
- This could probably also use a --with-fsplib or something to specify the
path, had problems with that too, because fsplib-0.7 installed into
/usr/local, ...
This changes the init target to be idempotent: most importantly it will now
never overwrite a Makefile if it exists. Additionally 'make init' will
generate the .vimrc files. Yay, no more stupid 'added fairies' commits! ;)
- Remove AC_PROG_MAKE_SET; we's be using GNU.
- Don't AC_SUBST() SPARSE; AC_PATH_PROGS() already does that.
- Fix cleaning after previous builds by moving up builddir setting
and only cleaning under $builddir/src/ so as not to launch infinite
reruns of configure.
Either set CONFIG_SM_SCRIPTING in features.conf or pass to ./configure the
option --disable-sm-scripting. Now scripting is also enabled when needed
and not only if some other scripting backend is enabled.
Remove some remnants of SEE scripting backend.
The following patch makes the configure script of ELinks ignore an
installed GNUTLS that is too old. If --with-gnutls was specified, this
results in a fatal error.
Tested with:
- Debian libgnutls11-dev 1.0.16-14, configure --with-gnutls
-> error at configure time
- Debian libgnutls-dev 1.2.9-2, configure --with-gnutls
-> compiles OK
... but only if Makefile.config exists. cleanall will cause make to enter
even SUBDIR-no dirs and clean OBJS-no files. This should fix the linking
problem report by zas when running:
./configure --enable-fastmem ; make ; \
./configure --enable-debug ; make
SEE is David Leonard's Simple Ecmascript Engine. The SEE scripting backend
is very raw and not tested very much. The idea was to see what kind of
creature SEE is (and contradict pasky's aired opinion that no new features
are added anymore ;).
echo 'function goto_url() { return 'localhost'; }' > ~/.elinks/hooks.js
and get local for maximum security ...
FYI: SEE is smaller than Spidermonkey but doesn't have the same kind of
data-driven interface, although it looks like it is possible to build that.
Involves prefixing with $(srcdir) to some of the build rule variables. For
the builddir we create Makefiles which simply include the srcdir Makefile.
Add list make rule to get list of Makefiles to generate (find will get it
wrong for builddirs nested in srcdir).
There are still a few minor issues like the file paths echoed during make
install ...
Finally, we are able to identify the builds more accurately. Now we check
if commit-id (cogito tool) is available and pass the output to vernum.c
compilation which then exports it as a build_id string.
The first line of --version is now a bit long, we should probably move the
build date to a second indented line or something.
The root makefile is converted as well as some leaf Makefiles. This
also brings in the required infrastructure and adjusts configure.in
appropriately.
I converted only makefiles containing no configurable stuff, since
that'll require more consideration yet.
