[Clfs-dev] Trunk ppc : a final update from my Xmas build

[Ken Moffat]

 Well, getting a working X took rather longer than I had expected
(box locked very shortly after the screen went black in 'startx', I
assumed - wrongly - that gcc was doing something wrong and built a
less-patched gcc to try that).  Turned out it was a bad
xf86-video-ati driver.  So, on with the show...

 I've now completed my desktop, and even flash and you-tube are
(mostly) working (gnash is good, but not perfect).  Apart from the
problem from running the module-init-tools testsuite (already fixed)
I found the following issues:

1. Glibc cannot be compiled with the branch_update-1 patch.  This
was not a surprise to me, the only piece of code it alters is in
nptl/sysdeps/powerpc/tcb-offsets.sym and I already knew it broke
ppc64.  In fact, that was why I wanted to look at ppc.  I also
looked at a recent glibc snapshot before I started the build, and
couldn't see this change in it.

 I propose to upload a -1A patch and change the books to use that (I
can hardly call it -2 when it doesn't contain any newer changes).
With so much aggravation on different platforms, particularly in
relation to identifying issues with branch update patches (nobody
else uses the exact same version, so we can get a lot of issues
specific to the revision we pulled), I think we should keep older
versions of the branch update patches around, so I don't intend to
delete -1.  If anybody then jumps up and says "I need it on xyz" we
can look at the detail.

2. The gcc testsuite fails horribly: instead of 3 unexpected
failures in gcc, and 1 in g++, I got 11951 in g++ (the order of the
testing has changed), 30340 in gcc, 198 in libgomp, 520 in
libmudflap, and the tests baild in libstdc++.  However, since then
it has coped with all the code I've thrown at it, so I tend to
assume that either there is something else needed in the basic
infrastructure to run the tests (rather like the way libgcc_s had
to be added), or else a minor change somewhere has totally borked
the testsuites.

3. We've now got flex-2.5.34.  My interest in this is close to zero,
so long as it works.  The ticket on LFS implies there is a lot of
breakage in other packages.  So far, I haven't seen that.  It did
fail to build libIDL, but in the end I have a simple suggested patch
(at bugzilla.gnome.org, bug 507344, attachment 102158) - it now
barfs on '%p ', complains about the next few '%' identifiers, and
fails to include any of the source file in the output.  Changing the
first item to '%pointer ' seems to make it work.  Maybe there are
other packages with similar breakage.  Certainly, I didn't have any
compilation problems with the following which all invoke flex, and I
haven't noticed any usage or run-time problems:
 kbd-1.13, iproute2-2.6.23, bc-1.06, libbonobo-2.20.1,
gcalctool-5.20.2.

 Oh well, maybe I can now get the book's version of ppc64 fully
working.  I think I'll add my necessary gcc patch (for tihe kernel)
to the book _before_ I try it again with the various package updates
that have gone in over the last couple of months.

ĸen
-- 
das eine Mal als Tragödie, das andere Mal als Farce