Re: value of $CXX when building libraries

[Phil Edwards <pedwards at disaster dot jaj dot com>]

[cc'ing the v3 list]


On Wed, Jan 10, 2001 at 12:40:19PM +0100, Gabriel Dos Reis wrote:
> Phil Edwards <pedwards@disaster.jaj.com> writes:
> | On Wed, Jan 10, 2001 at 09:22:24AM +0100, Gabriel Dos Reis wrote:
> | > 
> | > Well, now I'm wondering: which value should be used to compiler and
> | > library tests?  libstdc++/tests_flags.in is already doing flags
> | > computations; maybe we could start from there...
> | 
> | That's very true.  That script does nearly all the work already.  A makefile
> | rule could either call it, parse the results, and then use those values
> | directly; we wouldn't ever need a 'bin' directory.
> 
> Oh God, you're right.  I think 'tests_flags --compiler' actually gives
> the right answer (now that I'm awaken, I think there is no need for
> further parsing, am I again being dense?) 

Running the version generated as part of my more recent build, it looks
like --compiler will work for the include paths with no more parsing
(warning, horrible line wrapping ahead).  From in the v3 build directory:

% cat src/libstdc++.INC
-nostdinc++ -I/home/pme/src/unified/libstdc++-v3/include -I/home/pme/src/unified/libstdc++-v3/include/std -I/home/pme/src/unified/libstdc++-v3/include/c_std -I/home/pme/build/build-2001-01-10-i686/i686-pc-linux-gnu/libstdc++-v3/include -I/home/pme/src/unified/libstdc++-v3/libsupc++ -I/home/pme/build/build-2001-01-10-i686/i686-pc-linux-gnu/libstdc++-v3/libio -I/home/pme/src/unified/libstdc++-v3/libio -I/home/pme/src/unified/libstdc++-v3/libmath

% ./tests_flags --compiler $PWD ~/src/unified/*3
-nostdinc++ -I/home/pme/build/build-2001-01-10-i686/i686-pc-linux-gnu/libstdc++-v3/include -I/home/pme/src/unified/libstdc++-v3/include/std -I/home/pme/src/unified/libstdc++-v3/include/c_std -I/home/pme/src/unified/libstdc++-v3/include -I/home/pme/src/unified/libstdc++-v3/libsupc++ -I/home/pme/src/unified/libstdc++-v3/libio -I/home/pme/src/unified/libstdc++-v3/testsuite -I/home/pme/src/unified/libstdc++-v3/include/backward -I/home/pme/src/unified/libstdc++-v3/include/ext

We wouldn't necessarily need to create the libstdc++.INC file anymore
(Alexandre? Sanity check?); if I recall, doing so was creating problems
with directories with colons in their names.


The output from --built-library could be split up like we're currently doing
to provide all the libtool/compiler invocations.  (The things you can do
after twiddling IFS never cease to amaze me, even after all this time.)


> Per could you try that option?

BTW, Per, be sure to look at the comments at the start of tests_flags[.in]
if you try --built-library or --installed-library.  The results were driving
me loony until I read that they're /supposed/ to be colon-separated.  :-)


-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.