Re: gmp++ on multilib not compiling

Hi,

On Thu, 27 Jan 2022, Alain Toussaint wrote:

> I'm building a multilib enabled LFS system and gmp-6.2.1 won't build 
> for the 32-bit i686 part
I did build multilib LFS recently (January 9th) and did *not* run into 
trouble with gmp-6.2.1, but my build was manual and it was my first after
/usr-merge [So occassionally I may have fixed a typo or corrected 
some obvious problem to the (for me new) /usr dirs vs links on the fly].

> because the
> configure script does not find std iostream. error message in question:
> checking C++ compiler g++  -m32 -O2 -pedantic -fomit-frame-pointer 
> -mtune=generic -march=i686... no, std
> iostream, program does not run
> configure: error: C++ compiler not available, see config.log for details
Then most likely gmp was right that there really was no m32-g++ at hand...
For me the line in question ends in "yes".

I vaguelly remember some trouble with the m32-parts early on, lets see 
whether I can reconstruct anything that could be relevant for you... 
Checking a backup of my scripts tells me that I realized that the libstdc 
build shortly before does not only fill the current "build" dir as 
expected but also parallel "32" and "x32" dirs. The problem with that 
for me was that cleaning build dir and repeating correctly in case of 
error is not sufficient, because buggy stuff from previous attempt remains 
in parallel 32 and x32 dirs. Installing that would explain broken m32-g++ 
afterwards...

I also seem to remember that fetching something failed that surely was at 
the right place. Something along the lines of "x64 variant found out 
of the box or after ldconfig, but x32 part only fetched after re-login or 
re-mount". I don't remember more details, but I did have such "scripts 
unexpectly did not run in one go, but re-doing without script change 
worked" experience too - I suppose it was with freetype for x32. Anyway, 
I remembered that when you mentioned "could be an automation issue" 
idea...

> I found a github bug report for gmp which explain a possible fix:
> https://github.com/bazelbuild/rules_foreign_cc/issues/296 (even though 
> it is a bazel bug report, it is related to gmp).
I doubt this is related. I may not build strictly by the book, but 
especially the multilib part is surely *very* close to 
https://www.linuxfromscratch.org/~thomas/multilib/chapter08/gmp.html
- without trouble. I'd say: not gmps fault.

HTH!?
Uwe

-- 
http://lists.linuxfromscratch.org/sympa/info/lfs-dev
Unsubscribe: See the above information page