Discussion on Chap. 6 Glibc compile problem...

Roger Merchberger zmerch at 30below.com
Wed Jun 30 08:01:53 PDT 2004


I've been doing more reading on the compilation error that occurs on some 
AMD (and newer Intel CPUs) and also on my Crusoe processor during the make 
check of glibc in chapter 6... and there's a good reference to it in the 
Chapter 5 glibc section of 'da book':


Here is a list of the most common issues we are aware of:

   *  The math tests sometimes fail when running on systems where the CPU 
is not a relatively new genuine Intel or authentic AMD. Certain 
optimization settings are also known to be a factor here.

=-=-= and later =-=-=

The Glibc in Chapter 6 is the one we'll ultimately end up using, so that is 
the one we would really like to see pass the tests (but even there some 
failures could still occur -- the math tests, for example). When 
experiencing a failure, make a note of it, then continue by reissuing the 
make check. The test suite should pick up where it left off and continue. 
You can circumvent this stop-start sequence by issuing a make -k check. But 
if you do that, be sure to log the output so that you can later peruse the 
log file and examine the total number of failures.


So it sounds like if the glibc 'make check' fails only on the math section, 
it's still a good compile & one can continue normally.

Is there a way for the profile to use 'make -k check' as noted above and 
have nALFS 'track' how many errors there were in the testing phase (and/or 
which test[s] failed) [1]?

Or, as my father-in-law used to say "am I just full of condensed milk?" ;-)

Roger "Merch" Merchberger

[1] as I was typing this, I was re-running the make check, and the 
linuxthreads test also failed... hmmm... maybe it's not a good compile 
after all... :-(

Roger "Merch" Merchberger   | "Profile, don't speculate."
sysadmin, Iceberg Computers |     Daniel J. Bernstein
zmerch at 30below.com          |

More information about the alfs-discuss mailing list