mirror of
git://git.sv.gnu.org/autoconf
synced 2024-12-15 02:20:10 +08:00
e491f83180
including when the cache is used. There are still several issues: (i) there are too many runs of m4 (one for include, one for warnings, and some more), (ii) warnings spreading on several lines are not handled gracefully, (iii) the code meant to have the call stack display for errors does not work (its handling should move from m4 to autom4te). * bin/autom4te.in Autom4te::Channels, Autom4te::ChannelDefs): Use them. (@preselect): Add m4_warn. ($exit_status): Remove, use $exit_code. ($help): Use Autom4te::ChannelDefs::usage. (&handle_m4): No longer define the m4_warnings. At each run, extract and report the warnings. Always cache the result, including if the exit status is on failure, since if nothing changes, we should result in the same failure, hence we can use the cache. * lib/m4sugar/m4sugar.m4 (m4_warning_ifelse, _m4_warning_ifelse) (_m4_warning_error_ifelse, __m4_warning_error_ifelse, _m4_warn): Remove. (m4_warn): Redefine as a do-nothing: it is its invocation that matters, as warnings are now reported via traces. * lib/autoconf/general.m4 (AC_DIAGNOSE): Don't make it a copy of the contents of m4_warn: make it _call_ m4_warn, so that tracing the latter reveals calls to the former. Adjust the tests. * tests/m4sugar.at (m4@&t@_warn): Use existing warning categories. |
||
---|---|---|
.. | ||
.cvsignore | ||
m4sh.m4 | ||
m4sugar.m4 | ||
Makefile.am | ||
Makefile.in |