glibc 2.3.5 "make check" invalid multibyte sequence
 

During make check of a from-source build of glibc-2.3.5, I get the following:
:eek:

Configured with: Others have reported similar things before, on Google, but no one seemed to have a resolution for it.

The problem remains unresolved. And unfortunately, it stops make install at a rather awkward moment ... that is to say, unfinished!

Yes, it left me with a half-installed glibc. Not pretty. :cry: ... Recoverable, mind you, but not pretty!

Tested using {LANG=C, LC_ALL=C} and various other combinations. No effect.

I must say that I am quite puzzled why this would surface both on make check and make install.

i think the flag --enable-kernel=2.6.12
should be
--enable-kernel=2.6.0
and when you compile glibc 2.3.5 while running a kernel = or greater than 2.6.11
you need to use
glibc-2.3.5-fix_test-1.patch

http://www.linuxfromscratch.org/patc...x_test-1.patch

Nice idea... didn't work. :(

(directed here from an LFS post)

if possible repeat install of everything, exactly the same except have the only difference be glibc-2.3.4

This will at least help identify the problem down to glibc. if it fails identically with glibc-2.3.4, the problem is probably elsewhere considering how different those two glibc's are!

The problem appears to be more pervasive than I originally thought ... surfacing in several other compiles ... and this seems to point away from glibc and toward another tool, gettext.

Close inspection reveals that there is a very old copy of gettext on the system, and it is the one being picked-up in the path. I am removing it to see if the problem clears up, and will keep you informed.

Things are not going well. gettext-0.14.5 is failing its own test suite (make check) quite consistently, with many different language po's.

I'm going to presume that something must be wrong with the locale information (glibc) but right now I am seriously befoozled as to just what the source of this problem is. I do know that it seems to be highly pervasive, affecting many compiles of many different things, all with the same symptoms. I have removed the errant older copies of gettext and am reinstalling locales, but the problem is still open.

An example of the offending sequence is the two bytes: <B3><EA>. Hmmm... :scratch: .. :study: ..

what compiler are you doing this with ?
generally this stuff is working with gcc-3.4.4

It's 3.4.4. A clean copy of gettext, configured --with-included-gettext, is failing every time along with everything else.

Obviously the problem can't be the compiler .. it compiles just fine. It must be something off-the-wall that I don't understand yet, because it's affecting many things at the same time. What I call a "screwball problem." The worst to diagnose, although usually the simplest to solve once identified.

I see an unfortunate crash-course in GNU internationalization in my future. Until I fully understand what the check is saying to me, I can't seem to trace it back to its source.

Oh, this system is none-too-stable right now .. :cry: .. in many respects, but of course, I put it there, using a sacrificial system just as I advise so often, just so that I could figure my way back out. :rolleyes: ... But this problem is highly-pervasive, and curious.