egcs-ss-970821 - Run-Time Library Cross Problems

egcs-ss-970821 - Run-Time Library Cross Problems

[Joel Sherrill]

While compiling egcs-ss-970821 to target m68k-rtems, I had to disable
Objective-C and g77 because both assume the existence of a target
C library to build their runtimes.  As we add more and more language
front ends, the problems associated with target runtime files is getting
worse.  This seems to be increasing the pain of building a cross to an
embedded target.

THe objective-c runtime problem is well known.  The g77 error message was:

/usr1/rtems/work/build-m68k-tools/gcc/xgcc
-B/usr1/rtems/work/build-m68k-tools/gcc/ -c -g0 -o VersionI.o
/usr1/rtems/work/src/gcc/f/runtime/libI77/Version.c
/usr1/rtems/work/src/gcc/f/runtime/libI77/Version.c:265: stdio.h: No such
file or directory

For now, I have disabled both objective-c and fortran in my build tree.
I would like to include them though. :(

This does point out a benefit of including as many of the language
frontends as possible in egcs.  I am reporting problems with a cross
embedded configuration of g77. I certainly would never have turned up this
problem if it had not been in the tree. :)

--joel

New bugs in egcs

[H.J. Lu]

Hi,

Those extra failures don't exist in my gcc2 ss-970802.


-- 
H.J. Lu (hjl@gnu.ai.mit.edu)
--
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fomit-frame-pointer -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -funroll-all-loops -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fomit-frame-pointer -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -funroll-all-loops -pipe
ERROR: execute/920625-1.c: compiled program got signal 6: -O3 -funroll-all-loops -pipe
ERROR: 970731-1/runit: shell script returned error status: -O2 -fno-omit-frame-pointer -pipe
ERROR: 970731-1/runit: shell script returned error status: -O3 -fno-omit-frame-pointer -pipe
ERROR: 970731-1/runit: shell script returned error status: -O3 -funroll-all-loops -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -fpic -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fomit-frame-pointer -fpic -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -funroll-all-loops -fpic -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -fpic -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fomit-frame-pointer -fpic -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -funroll-all-loops -fpic -pipe
ERROR: execute/920625-1.c: compiled program got signal 6: -O3 -funroll-all-loops -fpic -pipe

Re: New bugs in egcs

[Jeffrey A Law]

  In message <m0x2mUF-0004ecC@ocean.lucon.org>you write:
  > But I was told that it was not a bug while I am keeping getting it. I
  > am afraid that egcs may be released with that cp/tree.c bug, which may
  > be specific to x86.
Ummm, what I remember reading was that your _patch_ was incorrect;
I don't think Jason (or anyone) said that it was not a bug.  He
asked you to verify that the bug still exists after the next snapshot.

Please do that with tonight's snapshot using the libstdc++ included
therein (no, I haven't made the snapshot yet, but the plan is to
get it done before I go to sleep tongiht).




jeff

Re: New bugs in egcs

[Jim Wilson]

	Great. Now only if I can compile libg++ with egcs. Is someone
	working on that cp/tree.c bug I reported?

I am looking into all bugs that have been reported (excepting C++ ones out
of my area of expertise).  Please be patient, and I will get to it as soon
as I can.

Jim

Re: New bugs in egcs

[H.J. Lu]

> 
> 	Great. Now only if I can compile libg++ with egcs. Is someone
> 	working on that cp/tree.c bug I reported?
> 
> I am looking into all bugs that have been reported (excepting C++ ones out

Thanks.

> of my area of expertise).  Please be patient, and I will get to it as soon
> as I can.
> 

But I was told that it was not a bug while I am keeping getting it. I
am afraid that egcs may be released with that cp/tree.c bug, which may
be specific to x86.

-- 
H.J. Lu (hjl@gnu.ai.mit.edu)

Re: New bugs in egcs

[H.J. Lu]

> 
> 	Those extra failures don't exist in my gcc2 ss-970802.
> 
> These problems all go away with my alias.c patch and jfc's unroll.c patch
> applied.
> 
> I get exactly the same c-torture results for egcs and gcc2, with egcs
> bootstrapped using -O6 -g.
> 

Great. Now only if I can compile libg++ with egcs. Is someone
working on that cp/tree.c bug I reported?

-- 
H.J. Lu (hjl@gnu.ai.mit.edu)

Re: New bugs in egcs

[Jim Wilson]

	Those extra failures don't exist in my gcc2 ss-970802.

These problems all go away with my alias.c patch and jfc's unroll.c patch
applied.

I get exactly the same c-torture results for egcs and gcc2, with egcs
bootstrapped using -O6 -g.

Jim

Re: New bugs in egcs

[H.J. Lu]

> 
> 
>   In message <m0x2JoO-0004ecC@ocean.lucon.org>you write:
>   > Hi,
>   > 
>   > Those extra failures don't exist in my gcc2 ss-970802.
> So, after re-building gcc with jfc's patch, then re-running the
> testsuite, how many of these new failures remain?
> 
> Jeff
> 
> 
> 

Here is the partial list. Somehow egcs screws up the real constant
numbers when the compiler is compiled with -O6 -g. I am looking into
it. I may not be able to find time before next weekend.

Also, that cp/tree.c change bug I reported earlier still exists,
although it is also in gcc2 ss-970802.

I also reported/fixed a few bugs and I have sent my patches to
gcc2. Should I resend them to egcs? My patches may look ugly. But
the bugs are real.


H.J.
-- 
H.J. Lu (hjl@gnu.ai.mit.edu)
---
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fomit-frame-pointer -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -funroll-all-loops -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fomit-frame-pointer -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -funroll-all-loops -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -fpic -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -fomit-frame-pointer -fpic -pipe
ERROR: execute/cvt-1.c: compiled program got signal 6: -O3 -funroll-all-loops -fpic -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fno-omit-frame-pointer -fpic -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -fomit-frame-pointer -fpic -pipe
ERROR: execute/conversion.c: compiled program got signal 6: -O3 -funroll-all-loops -fpic -pipe

Re: New bugs in egcs

[Jeffrey A Law]

  In message <m0x2JoO-0004ecC@ocean.lucon.org>you write:
  > Hi,
  > 
  > Those extra failures don't exist in my gcc2 ss-970802.
So, after re-building gcc with jfc's patch, then re-running the
testsuite, how many of these new failures remain?

Jeff