[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: problem with ./configure
From: |
Eric Blake |
Subject: |
Re: problem with ./configure |
Date: |
Thu, 12 Sep 2013 10:13:43 -0600 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130805 Thunderbird/17.0.8 |
On 09/12/2013 10:05 AM, Zack Weinberg wrote:
> On 2013-09-12 11:25 AM, Eric Blake wrote:
>>
>> -for ac_option in --version -v -V -qversion; do
>> +for ac_option in --version -v -V -qversion -version; do
>> _AC_DO_LIMIT([$ac_compiler $ac_option >&AS_MESSAGE_LOG_FD])
>
> This isn't anything to do with your change, but -qversion should be
> moved to the second position in this list, because the AIX compiler's
> response to any unrecognized command-line option is to dump its _entire
> manpage_ to stdout, which is IIRC about 10,000 lines of text. I see
> that _AC_DO_LIMIT clips to ten lines by default, so we're not gonna get
> a ridiculous amount of junk in config.log, but it would still be more
> efficient to have it closer to the beginning (I only suggest we put it
> second because it's far more likely that the user actually has a
> compiler that understands --version).
That could help if we had a 'break' statement to abort the loop once we
think a version had been scraped; but right now, the full loop is
executed to completion regardless of what option(s) the compiler
understands (and in the case of gcc, executing both --version AND -v is
useful, as they give different outputs, both of which are useful).
I'm okay if someone writes a patch to propose such an early exit (the
fewer times we execute 'gcc', the faster a configure script will run -
provided we aren't adding more forks in our decision for an early exit
in the first place), but am not sure I will tackle it myself.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature