Re: possibly undefined macro: dnl

[Akim Demaille]

>>>>> "adl" == Alexandre Duret-Lutz <address@hidden> writes:

adl> Instead, you could grep the generated configure for occurences of
adl> dnl and AC_LIBTOOL_CXX, it should be easier to spot the error.

There are several issues here I'd like to comment.

First of all, this scheme (looking in the source (note the singular)
for a possible origin) does not work properly for other Autom4te
worlds than the Autoconf one.

I.e., indeed, in Autoconf speaking, the last file, configure.ac, is
usually where the error is to be found, yet, we should sometimes look
for it in aclocal.m4 or even acinclude.m4 etc.

Now, have a look at the Autoconf test suite: it has a file suite.m4
which is just a set of m4_include.  In here the approach is useless.

So, I'm considering this scheme instead:

* Output the actual line where the culprit was found, i.e., not
  the _input_ file, but the _output_ file.

* Look for the culprit token in all the sources in reversed order
  and output it with an error message that clearly demonstrates that
  it's only a guess.

Technically, I not sure I already know how to browse the list of
inputs in their actual order (i.e., I'm able to get the list of
explicit inputs, and the list of includes, but I don't know yet how to
reorder them properly).

Also, I don't know how to display the hint message in such a way that
compile-mode could track it and make it possible to click on it.  I
should have a look at the GCC error msgs I guess.