Re: [libstdc++] Add --disable-hosted-libstdcxx, clean up conditionals

[Phil Edwards <phil at jaj dot com>]

On Thu, Aug 28, 2003 at 09:54:32PM +0200, Gabriel Dos Reis wrote:
> Bill Wendling <wendling@ncsa.uiuc.edu> writes:
> | Also sprach Gabriel Dos Reis:
> | } "Zack Weinberg" <zack@codesourcery.com> writes:
> | } 
> | } | Gabriel Dos Reis <gdr@integrable-solutions.net> writes:
> | } | 
> | } | Technical limitations of Bourne shell unfortunately have to trump
> | } | aesthetics.
> | } 
> | } Sure.  But, my motivation was not aesthetics.
> | } 
> | The "+" is not acceptable in a flag. It's a limitation, as mentioned
> | above. So the common way to deal with this is to replace + with x. It's
> | really that simple. This is a valid, well-thoughtout reason why you can't
> | have + in there.
> 
> Did I say the reason Phil enounciated was not valid?  No.  So
> side-tracking the discussion on that road won't be fruitful.

Gaby, we were genuinely wondering where you were going with this.
It's not sidetracking.


> "+" is not acceptable for a bourne shell variable name, fine.  I don't
> dispute that simple fact.  It was decided that that "+" should be
> replaced with "x". That also is fine.  And I don't disupte it.  
> That conversion may be automated by having sed translate "+" to "x".
> Should the user-specified be written with the machine oriented name?
> I think no.  And that is what motivated my request.
> the option --disable-hosted-libstdc++ flag can be translated to
> "disable_hosted_libstdcxx" variable name.  There is no innovation
> there. 

Aha!  An explanation!

That's not a bad idea, but it still won't work.  The variable names are
generated by autoconf, not by us.  Specifically, the problem cases of

    if test ${foo_libstdc+++set} = set; then

are all done by autoconf.

Yes, with sufficient sed wizardry, we could generate 'configure' and then
do some post-processing before checkin.  That's more trouble than it's
worth to save the user from typing "configure --help" and learning what
the spellings are, or from going to the web page where it's all documented.
It's also a nonstandard procedure for GCC's CVS, and one I would oppose.


> Are you also proposing we should say --enable-languages=cxx instead of
> the straight --enable-languages=c++?  Atre you also proposinig we
> rename the compiler from g++ to gxx?

Not the same thing.  None of those names have to be tested by autoconf.
(And we all know that there have been C++ compilers whose names were purely
alphanumeric for precisely this reason.  It's not uncommon.)


Phil

-- 
Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it.
    - Brian W. Kernighan