Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
[solved] flex-2.5.35+ test failure
View unanswered posts
View posts from last 24 hours

 
Reply to topic    Gentoo Forums Forum Index Portage & Programming
View previous topic :: View next topic  
Author Message
wcg
Guru
Guru


Joined: 06 Jan 2009
Posts: 588

PostPosted: Sun Aug 05, 2012 1:59 pm    Post subject: [solved] flex-2.5.35+ test failure Reply with quote

Flex-2.5.35, 2.5.35_p10, and 2.5.37 is failing a couple of tests in its test suite
on a core2 (xeon) platform (x86_64-pc-linux-gnu, no multilib).
Code:

kernel-3.3.8
linux-headers-3.4.0
glibc-2.15-r2
gcc-4.5.3-r2
CFLAGS="-march=native -O2 -fno-strict-aliasing -pipe -Wall"
LDFLAGS="${LDFLAGS} -Wl,--as-needed"
FEATURES="splitdebug"

[edit]
Code:

CFLAGS="-march=native -O2 -fno-strict-aliasing -ggdb -pipe -Wall"

[/edit]
No exotic USE flags for flex. (I noticed that Stage2 and Stage3 did not
match in the gcc-4.5.3-r2 build log, too. Just fyi, since there could be
a zillion possible reasons for that, and it still might be the most stable gcc
version available on most CHOST platforms even with the error.)

The two tests that are failing are
Code:

test-bison-yylloc
test-bison-yyllval

(No bison-2.6.2 test suite failures.)

I can reproduce the test failures on demand if anyone needs
to see the test output in detail, and I still have the emerge logs
for all three flex versions.

I do not see any flex (2.5.35_p10) test failures on an amd64 running on
kernel 3.2.12. (The only system I really have to compare with at
the moment.) I will update the amd system to kernel 3.3.8 within a week,
If the same flex test failure turns up there I will report it in this thread.
_________________
TIA


Last edited by wcg on Wed Aug 15, 2012 4:57 am; edited 1 time in total
Back to top
View user's profile Send private message
wcg
Guru
Guru


Joined: 06 Jan 2009
Posts: 588

PostPosted: Mon Aug 06, 2012 3:30 pm    Post subject: Reply with quote

PS:
binutils--2.22.90

emerge --info:
http://pastebin.ca/2177962
_________________
TIA
Back to top
View user's profile Send private message
wcg
Guru
Guru


Joined: 06 Jan 2009
Posts: 588

PostPosted: Mon Aug 06, 2012 5:07 pm    Post subject: Reply with quote

So, I updated the kernel on the amd64 box to 3.3.8,
along with a few other tools. flex-2.5.37 completed
make check with no test failures.

Then I updated binutils from binutils-2.22-r1 to
binutils-2.22.90, and the flex test suite failed
the same two tests, test-bison-yyloc and test-bison-yylval.

So it seems that flex-2.5.35+ and binutils-2.22.90 have some
details to work out, and sticking with 2.22-r1 until the issue
is identified would be more stable (at least on my x86_64-pc-linux-gnu
systems).
_________________
TIA
Back to top
View user's profile Send private message
toralf
Advocate
Advocate


Joined: 01 Feb 2004
Posts: 2980
Location: Hamburg/Germany

PostPosted: Mon Aug 06, 2012 7:07 pm    Post subject: Reply with quote

You already checked b.g.o , eg. https://bugs.gentoo.org/show_bug.cgi?id=429954 ?
Back to top
View user's profile Send private message
wcg
Guru
Guru


Joined: 06 Jan 2009
Posts: 588

PostPosted: Mon Aug 06, 2012 11:06 pm    Post subject: Reply with quote

No, I had not checked that bugreport URL, but that is the same
error.

On the amd64 system, the whole build chain was compiled with
gcc-4.5.3 (probably -r1 for most of it; maybe -r2 for some things
that had been updated in the last few months) and assembled
and linked with binutils-2.22-r1. flex did not show the errors
there until updating binutils to binutis-2.22.90.

On the core2 system, I tried backing off to binutils-2.21.1-r1,
the last version of binutils marked stable for amd64, and
re-emerging flex-2.5.37 (with FEATURES="test ..."). The error
persists. I tried re-emerging gcc-4.5.3-r2 using the stable
binutils and re-emerging flex, I still get the error.

So it is not as simple as "flex built with a new, unstable binutils".
It is not even as "simple" as gcc-4.5.3 built with binutils-2.22.90
(instead of an older binutils) and then compiling flex. The problem
seems to be some other obscure tool or environment variable
setting in the build chain. binutils-2.22.90 exposed it, but it
is not necessarily the cause of flex straying from its usual
predictable behavior.

So, I am working through the build tools and older flex versions to
find at what point the flex test suite no longer errors out on the core2
system. (On the amd64 system, almost everything is still built
with the older binutils-2.22-r1 that did not trigger the flex test
failure.)
_________________
TIA
Back to top
View user's profile Send private message
wcg
Guru
Guru


Joined: 06 Jan 2009
Posts: 588

PostPosted: Tue Aug 07, 2012 2:50 pm    Post subject: Reply with quote

After rebuilding all of the build tools on the core2 system with
binutils-2.21.1-r1, I still get the error from the flex (2.5.35+) test suite.
One thing that both systems had in common was updating to
linux-headers-3.4-r2 at the same time as the binutils update
to binutils-2.22.90.

For the core2 system, this was a new install, so the previous linux-headers
is whatever the install-amd64-minimal-20120621 install cd and stage3
tarball installed. On the amd system, the previous linux-headers was
linux-headers-3.1. glibc, util-linux, iproute2, and a variety of other
packages that depend on linux-headers were re-emerged to take
into account any changes to linux-headers between 3.1 and 3.4.

This was done to accomodate the upgrade to kernel-3.3.8 on
both systems. (Likely the error results from something that glibc
does differently when compiled against the updated kernel-headers
that the flex code does not expect.)

The concurrency of the binutils update with the appearance
of the flex error could have been coincidental and unrelated to
the change in flex's behavior when running the "make check"
test suite that it runs with "FEATURES=test".
_________________
TIA
Back to top
View user's profile Send private message
wcg
Guru
Guru


Joined: 06 Jan 2009
Posts: 588

PostPosted: Wed Aug 15, 2012 4:57 am    Post subject: Reply with quote

Solved: according to the bugzilla bug report page cited above,
the problem turned out to be a "flex test suite running with
bison-2.6+" problem (not binutils and not linux-headers-3.4.0-r2).
_________________
TIA
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Portage & Programming All times are GMT
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum