[gentoo-amd64] Re: Is -fPIC compiler sensitive?

[Duncan]

Duncan posted <pan.2006.02.15.07.53.34.26158@×××.net>, excerpted below, 
on Wed, 15 Feb 2006 00:53:34 -0700:

> Simon Stelling posted <43F23F8A.80905@g.o>, excerpted below,  on
> Tue, 14 Feb 2006 21:37:30 +0100:
>
>> Duncan wrote:
>>> Is -fPIC gcc-version sensitive?
>>
>> It might be that some version of gcc prints a warning when configure
>> performs a test to decide whether -fPIC is a supported flag or not and
>> another version doesn't. Depending on that, configure might decide that
>> it's not supported and compile everything without -fPIC. Have a close
>> look to configure's output and config.log, I'm pretty sure that's the
>> issue here.
>
> Thanks.  I'll check.

FWIW, I did check, and you were right about the warning, but it wasn't
due to -fPIC, but rather to one of my CFLAGS/CXXFLAGS (the warning:
-freorder-blocks-and-partition does not work with exceptions).  gcc-3.4.x
doesn't print that warning, taking the flag without protest (tho it
probably still ignores it, don't care enough to go find the gcc-3.4 docs
and look), the gcc-4.1 snapshots do.

So...  The configure script simply interpreted the warning about the cflag
being ignored as a warning about its added -fPIC being ignored.  It
therefore compiled without -fPIC and naturally failed.

The problems with -freorder-blocks-and-partition do seem to be mostly with
c++ code (as this was, the regular c test earlier in the script worked
just fine), but also seem to be more trouble than they are worth. I'm
thinking I'll now have to split them up, tracking them separately, much as
I hate to do so. (The weaker non-partitioning version -freorder-blocks,
works just fine, as I've had it in my CFLAGS as well, since the long
cflags post I made some weeks ago, without issue.  The problems are all
with the -and-partition version, and almost all with c++, not c.)

Hmm...  I wonder which warning option controls those warnings.  Maybe I
can turn it off in the same C(XX)?FLAGS variables that turn on the flag
creating them?  Since it seems to be mainly a c++ problem, however, little
if any C++ code appears to be able to use the partition thing anyway, and
I haven't investigated gcc's warning options much, I think I'll try the
separate CXXFLAGS without that for awhile, first, tho.

I'm actually quite relieved, as I had noticed a few more -fPIC errors
recently, apparently all due to this.  I was beginning to worry that all
the -fPIC work that the Gentoo amd64 team has done was somehow getting
lost!  Finding out it was my own system is therefore /quite/ a relief, as
I'd hate to have to go thru all those porting struggles again!  I'm sure
you'll agree!  =8^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@g.o mailing list