> > I checked that bin/autoconf, tests/testsuite, and configure > > all get the correct number and location of as_fn_error definitions (and > it > > wasn't very easy, since I intentionally made the shell function > definition > > dependent on whether logging is in use, so that bin/autoconf's version > is > > lighter-weight than configure's version). > > I appreciate the effort, but is this really necessary? How much do you > save?
Oh, how about all of _AS_LINENO_PREPARE (which in turn drags in _AS_CR_PREPARE, changes as_detect_suggested to a non-empty string, ...)? It makes a huge difference to bin/autoconf - we really _don't_ want to slow down the wrapper script to re-exec itself with a LINENO-aware shell just to spawn off a perl script. Yet that is what happens if as_fn_error blindly uses _AS_ECHO_LOG rather than waiting until m4_wrap time to see if a log was actually used. > ps: I did read the patch. :-) OK, then I've committed it. -- Eric Blake -- View this message in context: http://www.nabble.com/shell-function-for-AS_ERROR-tp20627146p20629390.html Sent from the Gnu - Autoconf - Patches mailing list archive at Nabble.com.
