egcs-1.0 test results on i586-pc-linux-gnulibc1

egcs-1.0 test results on i586-pc-linux-gnulibc1

[Horst von Brand]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1512 bytes --]

linux-2.1.71, libc-5.4.41, binutils-2.8.1.0.17;
--enable-haifa --enable-shared; -O2 -fomit-frame-pointer -march=pentium
(<asm/signal.h> patched to provide NSIG)

WARNING: Couldn't find the global config file.
Test Run By root on Fri Dec  5 01:45:28 1997
Native configuration is i586-pc-linux-gnulibc1

		=== libio tests ===

FAIL: tfformat.c compilation 
FAIL: tiformat.c compilation 
FAIL: tstdiomisc.c compilation 

		=== libio Summary ===

# of expected passes		33
# of unexpected failures	3
# of expected failures		4

		=== libstdc++ tests ===

		=== libstdc++ Summary ===

# of expected passes		30

		=== gcc tests ===

		=== gcc Summary ===

# of expected passes		4883
# of expected failures		5
# of unsupported tests		7

		=== g++ tests ===

XPASS: g++.jason/destruct3.C - (test for bogus messages, line 38)
XPASS: g++.mike/dyncast1.C  Execution test
XPASS: g++.mike/dyncast2.C  Execution test

		=== g++ Summary ===

# of expected passes		3400
# of unexpected successes	3
# of expected failures		80
# of untested testcases		6

		=== g77 tests ===

FAIL: g77.f-torture/execute/dnrm2.f execution,  -O2 -fomit-frame-pointer -finline-functions -funroll-loops 
FAIL: g77.f-torture/execute/dnrm2.f execution,  -O2 -fomit-frame-pointer -finline-functions -funroll-all-loops 

		=== g77 Summary ===

# of expected passes		130
# of unexpected failures	2
-- 
Horst von Brand                             vonbrand@sleipnir.valparaiso.cl
Casilla 9G, Viña del Mar, Chile                               +56 32 672616

linux 2.1.71 and libc

[H.J. Lu]

> 
> linux-2.1.71, libc-5.4.41, binutils-2.8.1.0.17;

There is a problem with linux-2.1.71 and libc 5.4.41. I haven't
found a good solution for it yet.

sigset_t is changed in linux-2.1.71 to 64 bits. That breaks libc 5.4.x.
Also many older binaries using 32 bit sigset_t may not work properly
under linux-2.1.71.

It happens to glibc 2 also. The latest glibc 2.1 also has 64-bit
sigset_t. As the result, old binaries won't work very well. It may
be easier to deal with in glibc 2 using ELF versioning. But it is
not so for libc 5.4.x. For libc 5.4.x, I may do

extern xxxx _new_sigxxxx (xxx);

#define sigxxxx _new_sigxxxx

Ulrich, could you please fix it for glibc 2.1?

BTW, we have to translate between old and new kernel/sigset_t at
the run time. It is a very tricky one.

Linus, I think we shoud provide a new system call for each old system
call which uses sigset_t. Otherwise, we may run into many trouble
with old binaries.


Thanks.


H.J.

Re: linux 2.1.71 and libc

[Ulrich Drepper]

hjl@lucon.org (H.J. Lu) writes:

> It happens to glibc 2 also. The latest glibc 2.1 also has 64-bit
> sigset_t.

No.  We had and have a 1024 bit sigset_t.

> As the result, old binaries won't work very well. It may
> be easier to deal with in glibc 2 using ELF versioning.

What do you tell?  There is not problem like this since at user-level
we use a completely different type.

-- Uli
---------------.      drepper at gnu.org  ,-.   Rubensstrasse 5
Ulrich Drepper  \    ,-------------------'   \  76149 Karlsruhe/Germany
Cygnus Solutions `--' drepper at cygnus.com   `------------------------

Re: linux 2.1.71 and libc

[Horst von Brand]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1193 bytes --]

hjl@lucon.org (H.J. Lu) said:
> "Horst von Brand" at Dec 5, 97 07:10:52 am said:
> > linux-2.1.71, libc-5.4.41, binutils-2.8.1.0.17;

> There is a problem with linux-2.1.71 and libc 5.4.41. I haven't
> found a good solution for it yet.

> sigset_t is changed in linux-2.1.71 to 64 bits. That breaks libc 5.4.x.
> Also many older binaries using 32 bit sigset_t may not work properly
> under linux-2.1.71.

Hum.... I had to take a look at this (I needed to compile modutils-2.1.55,
but the compilation blew up due to incorrect types). The fix for this is
simple: In /usr/include/signal.h there are several #defines for the sigset
manipulation functions that assume 32 bits (or an unsigned long, at least),
they are at the start of an #ifdef __OPTIMIZE__ Just move the #ifdef 0 a
couple of lines lower up, and they aren't included. This will need a closer
look...

OTOH, I took a look at the whole stuff in 2.1.71, and I can't see anything
that might break 32-signal applications in there. If it did, my system
wouldn't work, I guess...
-- 
Horst von Brand                             vonbrand@sleipnir.valparaiso.cl
Casilla 9G, Viña del Mar, Chile                               +56 32 672616