[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: autoconf will accept a nonexistent compiler as the second one checke
Re: autoconf will accept a nonexistent compiler as the second one checked
Mon, 20 Mar 2006 17:02:52 +0100
* Martin Michlmayr wrote on Mon, Mar 20, 2006 at 02:15:34PM CET:
> * Stepan Kasal <address@hidden> [2006-03-20 12:00]:
> > Well, ./configure _could_ do more sanity checks, but the expense is
> > obvious: all the people who pass correct parameters would have to
> > wait another few seconds.
> Well, I think in this case some more sanity checks are justified
> because the few seconds lost checking if CXX exists are certainly
> better than all the headache caused by a configure script that "works"
> when CXX is bogus or that later fails at some completely unrelated
> place, such as in http://bugs.debian.org/357369 - where the error was:
> configure:20113: g++-4.0 -c -O2 -D_REENTRANT -fpermissive conftest.cc >&5
> ./configure: line 20114: g++-4.0: command not found
> This is really not a very nice (an obvious) way to fail.
If this is changed in Autoconf, it should be noted that Libtool-1.5.x
using packages will then fail hard unless a C++ and Fortran 77 compiler
are installed, even if the package in question does not use these
languages. This has only been fixed in the CVS version of Libtool,
This is not to advocate against the change, merely to point out this
issue; maybe I should look again at backporting that fix.. :-/