>> I am a leeetle surprised that it is a bug without the rest of my patch
>> -- this suggests that "eval" is even bugger than I thought.
>
> I don't understand this sentence.  If, by "bugger" you mean buggier,

Yes, sorry.

> then I wonder which "bug" you are referring to:

I was speculating at the time -- perhaps someone set IFS=' ' or
something like that.  But going back and reading
<http://lists.gnu.org/archive/html/autoconf-patches/2006-05/msg00109.html>
it appears that it may have just been for nicer config.log output; in
that case my fix wouldn't solve the problem.  So I'll do something
like this instead: it should fix Debian's problem too, whatever it
was.

-        AC_TRY_COMMAND([test -z "$ac_[]_AC_LANG_ABBREV[]_werror_flag"
-                        || test ! -s conftest.err]) &&
+        AC_TRY_COMMAND([test -z "$ac_[]_AC_LANG_ABBREV[]_werror_flag" || test 
! -s conftest.err]) &&


> But I probably have just misunderstood the above, and you meant bugs
> in Autoconf's use of "eval".

Yes, that's what the new documentation is about.


Reply via email to