This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: rfc: new libgcc build mechanism
- To: Phil Edwards <pedwards at jaj dot com>
- Subject: Re: rfc: new libgcc build mechanism
- From: pausmith at nortelnetworks dot com (Paul D. Smith)
- Date: 10 May 2000 15:33:07 -0400
- Cc: "Smith, Paul " <pausmith at BayNetworks dot COM>, autoconf at gnu dot org, gcc at gcc dot gnu dot org
- References: <200005101826.OAA05377@jaj.com>
- Reply-To: pausmith at nortelnetworks dot com
%% Phil Edwards <pedwards@jaj.com> writes:
Rats, the last one got away from me. Please ignore it.
>> I haven't looked closely at this, but I will note right off that this:
dt> "CC=$$(echo $$(case '$(CC)' in (stage*) echo '$(CC)' | sed -e
>> is very non-portable.
pe> I have looked at this even less closely than you; perhaps I have
pe> missed something. If you are constructing the makefile via
pe> autoconf, why not detect a shell that will do such expansions and
pe> set SHELL in the output makefile accordingly?
If you can find one, that's fine. First you are assuming that every
platform you want to build GCC on _will_ have such a shell. Since ksh
isn't free, and certainly not every UNIX includes bash, some versions of
UNIX might not have one anywhere (is this the same thread where we were
discussing the limitations of 4.3BSD sh?)
Second, you are assuming we can find them; UNIX vendors are notoriously
sneaky about where they stow these things, and also what they name
them. I don't relish the thought of providing a long list of potential
hiding places and aliases for shells within autoconf :-/.
Third, testing for all these features would be a pain. Not a huge pain,
but a pain nonetheless.
I still believe we can do what we need to do without these features. I
provided an implementation of the example given which I believe DTRT for
all shells... no?
pe> Also, just FYI, the default make on Solaris (from the development
pe> tools package, sitting in /usr/ccs/bin) is the same make as the
pe> POSIX-complaint make, so it would pay attention to the SHELL
pe> setting:
Sure, that's not the problem. I've never heard of _ANY_ make that
didn't obey the SHELL variable.
--
-------------------------------------------------------------------------------
Paul D. Smith <psmith@baynetworks.com> Network Management Development
"Please remain calm...I may be mad, but I am a professional." --Mad Scientist
-------------------------------------------------------------------------------
These are my opinions---Nortel Networks takes no responsibility for them.