* Exhaustive simulator testing
@ 2002-09-26 14:51 Zack Weinberg
2002-09-27 3:11 ` Richard Earnshaw
` (2 more replies)
0 siblings, 3 replies; 9+ messages in thread
From: Zack Weinberg @ 2002-09-26 14:51 UTC (permalink / raw)
To: gcc
Mostly to see what would happen, I wrote a script that built and
tested the combined tree for every last target in the matrix I posted
yesterday as part of the simtest-howto rewrite.
Just over half of them fail to build for various reasons:
d10v-elf - not supported by GCC
mn10200-elf - not supported by GCC (obsoleted target)
Did the d10v back end just never get checked in? It isn't mentioned
anywhere in gcc/ChangeLog, but I'm pretty sure I've seen reference to
its existing.
d30v-elf - not supported by GDB
fr30-elf - not supported by GDB
i960-coff - not supported by GDB
It Would Be Nice if it were possible to build the simulator without
also building GDB. It Would Be Really Nice if you didn't have to
have GDB's sources in the tree at all.
m32r-elf - ICE during build
newlib/libc/stdlib/dtoa.c: In function `_dtoa_r':
newlib/libc/stdlib/dtoa.c:854: internal compiler error: RTL flag check:
CONSTANT_POOL_ADDRESS_P used with unexpected rtx code `const' in
addr24_operand, at config/m32r/m32r.c:603
m6811-elf - libgloss bug
libgloss/m68hc11/syscalls.c:53: error: conflicting types for `sbrk'
newlib/libc/include/sys/unistd.h:97: error: previous declaration of `sbrk'
mcore-elf - ICE during build
libgcc2.c: In function `__lshrdi3':
libgcc2.c:265: error: Too many outgoing branch edges from bb 1
libgcc2.c:265: internal compiler error: verify_flow_info failed
mn10300-elf - GAS barfs during build
_muldi3.s: Assembler messages:
_muldi3.s:829: Error: can't resolve `.Letext0' {.text section}
- `.Ltext0' {.text section}
sparclite-elf - hits #error in crtstuff.c during build
<command line>:12:10: warning: "cpu" re-asserted
<command line>:13:14: warning: "machine" re-asserted
crtstuff.c:415:2: #error "What are you doing with crtstuff.c, then?"
crtstuff.c:183: warning: `__DTOR_LIST__' defined but not used
crtstuff.c:195: warning: `__EH_FRAME_BEGIN__' defined but not used
The rest do build, but the test suites are in varying degrees of
shape. Here's a matrix - I foolishly forgot to check out all the
language runtime libraries, so I only have meaningful results for C.
All these targets successfully built the C++, ObjC, and Fortran front
ends. (I'm not about to try to nail down which of these targets
support Java and Ada.)
arm-elf pass fail xpass xfail unres unsup untest
binutils 31 1
gas 63
ld 29 3
newlib 5 1
gdb 6832 291 51 30 11 9
gcc 20211 26 1 63 42 150
h8300-hms pass fail xpass xfail unres unsup untest
binutils 21 1 4
gas 81 3
ld 14 3 3
newlib 6
gdb 947 2150 34 197 9 9
gcc 11039 4593 2 54 4411 159
mips-elf pass fail xpass xfail unres unsup untest
binutils 30 2
gas 123 5 1
ld 42 1 4 1
newlib 5 1
gdb 6757 370 51 50 11 9
gcc 20165 67 86 52 148
mips64-elf pass fail xpass xfail unres unsup untest
binutils 30 2
gas 123 5 1
ld 41 2 4 1
newlib 6
gdb 4483 1737 2 140 86 11 9
gcc 20162 65 92 53 147
powerpc-eabisim pass fail xpass xfail unres unsup untest
binutils 31 1
gas 27
ld 24 2 1
newlib 6
gdb 6446 291 7 58 30 11 9
gcc 20258 22 62 42 137
sh-hms pass fail xpass xfail unres unsup untest
binutils 24 2
gas 28 2 2
ld 14 5 2 3 1
newlib 4 2
gdb 6174 436 2 147 32 11 9
gcc 20151 30 62 11 156
sparc64-elf pass fail xpass xfail unres unsup untest
binutils 25 7
gas 49 2
ld 19 13
newlib 6
gdb 1998 2039 40 133 7 9
gcc 660 14962 33 20 4496 142
v850-elf pass fail xpass xfail unres unsup untest
binutils 31 1
gas 47
ld 29 3
newlib 5 1
gdb 6695 285 90 59 30 11 9
gcc 20187 56 1 61 48 151
zw
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-09-26 14:51 Exhaustive simulator testing Zack Weinberg
@ 2002-09-27 3:11 ` Richard Earnshaw
2002-09-27 10:16 ` Tom Tromey
2002-10-02 7:29 ` Jeff Law
2002-10-18 6:53 ` Ben Elliston
2 siblings, 1 reply; 9+ messages in thread
From: Richard Earnshaw @ 2002-09-27 3:11 UTC (permalink / raw)
To: Zack Weinberg; +Cc: gcc, Richard.Earnshaw
> Mostly to see what would happen, I wrote a script that built and
> tested the combined tree for every last target in the matrix I posted
> yesterday as part of the simtest-howto rewrite.
> arm-elf pass fail xpass xfail unres unsup untest
> binutils 31 1
> gas 63
> ld 29 3
> newlib 5 1
> gdb 6832 291 51 30 11 9
> gcc 20211 26 1 63 42 150
He He: here's a more complete list from my own runs, including some
multi-lib varaints...
All tests run from a compiler checked out at: Thu Sep 26 08:44:01 UTC 2002
configuration: arm-elf
arm-sim
pass fail xpass xfail unres unsup untest
binutils 31 1
gas 64
ld 29 3
newlib[1] 5 1
gdb[2] 6447 264 64 493 7
gcc 20662 26 1 63 42 151
g++ 7463 15 93 21 11 29
g77 1564 43 31 9 8
objc 1153
libjava[3] 2034 18 14 104
libstdc++ tests not run -- ?? couldn't locate runtest
arm-sim/-mhard-float
pass fail xpass xfail unres unsup untest
binutils 31 1
gas 64
ld 29 3
newlib[1] 5 1
gdb[2] 6238 276 64 596 7
gcc 20534 154 64 42 151
g++ 7463 15 93 21 11 29
g77 1559 47 21 9 8
objc 1153
libjava[3] 2034 18 14 104
libstdc++ tests not run -- ?? couldn't locate runtest
arm-sim/-mthumb
pass fail xpass xfail unres unsup untest
binutils 31 1
gas 64
ld 29 3
newlib[1] 5 1
gdb[2] 6280 390 64 466 7
gcc 20649 36 2 62 44 151
g++ 7491 58 93 21 11 29
g77 1560 46 21 9 8
objc 1153
libjava[3] 2035 18 14 103
libstdc++ tests not run -- ?? couldn't locate runtest
Notes:
[1] Newlib tests are run in every multi-lib variant on every pass;
typically all tests will fail if the multi-lib does not match the test
config ;-(
[2] My gdb sources contain major changes from the mainline, so are not
really representative.
[3] Libjava requires some minor changes to enable the package for arm-elf.
R.
PS. Does your script also produce the above summary tables? Something to
do that automatically would be very cool.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-09-27 3:11 ` Richard Earnshaw
@ 2002-09-27 10:16 ` Tom Tromey
2002-09-28 7:16 ` Richard Earnshaw
0 siblings, 1 reply; 9+ messages in thread
From: Tom Tromey @ 2002-09-27 10:16 UTC (permalink / raw)
To: Richard.Earnshaw; +Cc: gcc
>>>>> "Richard" == Richard Earnshaw <rearnsha@arm.com> writes:
Richard> [3] Libjava requires some minor changes to enable the package
Richard> for arm-elf.
Are these changes suitable for checkin?
Tom
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-09-27 10:16 ` Tom Tromey
@ 2002-09-28 7:16 ` Richard Earnshaw
0 siblings, 0 replies; 9+ messages in thread
From: Richard Earnshaw @ 2002-09-28 7:16 UTC (permalink / raw)
To: tromey; +Cc: Richard.Earnshaw, gcc
> >>>>> "Richard" == Richard Earnshaw <rearnsha@arm.com> writes:
>
> Richard> [3] Libjava requires some minor changes to enable the package
> Richard> for arm-elf.
>
> Are these changes suitable for checkin?
>
> Tom
Yep, indeed I'd been planning to check them in today before I read your
mail.
R.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-09-26 14:51 Exhaustive simulator testing Zack Weinberg
2002-09-27 3:11 ` Richard Earnshaw
@ 2002-10-02 7:29 ` Jeff Law
2002-10-02 14:00 ` Alexandre Oliva
2002-10-18 6:53 ` Ben Elliston
2 siblings, 1 reply; 9+ messages in thread
From: Jeff Law @ 2002-10-02 7:29 UTC (permalink / raw)
To: Zack Weinberg; +Cc: gcc
In message <20020926214630.GG12223@codesourcery.com>, Zack Weinberg writes:
>Mostly to see what would happen, I wrote a script that built and
>tested the combined tree for every last target in the matrix I posted
>yesterday as part of the simtest-howto rewrite.
>
>Just over half of them fail to build for various reasons:
>
>d10v-elf - not supported by GCC
>mn10200-elf - not supported by GCC (obsoleted target)
Note that the mn102 may need to be resurrected. I'm hearing some rumblings
that it is being used inside Matsushita. Luckily it shouldn't be in
terrible shape. I built it a couple months ago and fixed a handful of
failures.
>mn10300-elf - GAS barfs during build
>
>_muldi3.s: Assembler messages:
>_muldi3.s:829: Error: can't resolve `.Letext0' {.text section}
> - `.Ltext0' {.text section}
Alex -- can you look at this?
Jeff
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-10-02 7:29 ` Jeff Law
@ 2002-10-02 14:00 ` Alexandre Oliva
0 siblings, 0 replies; 9+ messages in thread
From: Alexandre Oliva @ 2002-10-02 14:00 UTC (permalink / raw)
To: law; +Cc: Zack Weinberg, gcc
On Oct 2, 2002, Jeff Law <law@porcupine.cygnus.com> wrote:
>> _muldi3.s: Assembler messages:
>> _muldi3.s:829: Error: can't resolve `.Letext0' {.text section}
>> - `.Ltext0' {.text section}
> Alex -- can you look at this?
Yeah, it's been in my to-do list for a while already. I suspect an
assembler problem, since the assembly output looks right.
--
Alexandre Oliva Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist Professional serial bug killer
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-09-26 14:51 Exhaustive simulator testing Zack Weinberg
2002-09-27 3:11 ` Richard Earnshaw
2002-10-02 7:29 ` Jeff Law
@ 2002-10-18 6:53 ` Ben Elliston
2002-10-30 14:21 ` Zack Weinberg
2 siblings, 1 reply; 9+ messages in thread
From: Ben Elliston @ 2002-10-18 6:53 UTC (permalink / raw)
To: gcc
>>>>> "Zack" == Zack Weinberg <zack@codesourcery.com> writes:
Zack> It Would Be Nice if it were possible to build the simulator
Zack> without also building GDB. It Would Be Really Nice if you
Zack> didn't have to have GDB's sources in the tree at all.
Are you sure? Most simulators can be built without having to first
build GDB -- you just configure src/sim (or src/sid, if you're lucky
enough) and it will produce a standalone <target>-run executable.
Ben
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-10-18 6:53 ` Ben Elliston
@ 2002-10-30 14:21 ` Zack Weinberg
2002-10-31 8:54 ` Phil Edwards
0 siblings, 1 reply; 9+ messages in thread
From: Zack Weinberg @ 2002-10-30 14:21 UTC (permalink / raw)
To: Ben Elliston; +Cc: gcc
On Fri, Oct 18, 2002 at 09:52:44PM +1000, Ben Elliston wrote:
> >>>>> "Zack" == Zack Weinberg <zack@codesourcery.com> writes:
>
> Zack> It Would Be Nice if it were possible to build the simulator
> Zack> without also building GDB. It Would Be Really Nice if you
> Zack> didn't have to have GDB's sources in the tree at all.
>
> Are you sure? Most simulators can be built without having to first
> build GDB -- you just configure src/sim (or src/sid, if you're lucky
> enough) and it will produce a standalone <target>-run executable.
Many of the src/sim simulators make direct reference to GDB routines.
For instance, sim/ppc/main.c and sim/ppc/sim_calls.c both include
gdb/callback.h and gdb/remote-sim.h. If the GDB sources are not in
the tree these files will obviously not compile. My experience is
that if GDB is in the tree but hasn't been configured, they still
won't compile (bombs out looking for gdb/config.h or similar).
I haven't looked at sid.
zw
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: Exhaustive simulator testing
2002-10-30 14:21 ` Zack Weinberg
@ 2002-10-31 8:54 ` Phil Edwards
0 siblings, 0 replies; 9+ messages in thread
From: Phil Edwards @ 2002-10-31 8:54 UTC (permalink / raw)
To: Zack Weinberg; +Cc: Ben Elliston, gcc
On Wed, Oct 30, 2002 at 11:32:50AM -0800, Zack Weinberg wrote:
>
> Many of the src/sim simulators make direct reference to GDB routines.
> For instance, sim/ppc/main.c and sim/ppc/sim_calls.c both include
> gdb/callback.h and gdb/remote-sim.h. If the GDB sources are not in
> the tree these files will obviously not compile. My experience is
> that if GDB is in the tree but hasn't been configured, they still
> won't compile (bombs out looking for gdb/config.h or similar).
Yes. When building a combined tree for cross-targets, I have to configure
all of it, then delete gdb/Makefile -- and nothing else -- to avoid building
a debugger.
Phil
--
I would therefore like to posit that computing's central challenge, viz. "How
not to make a mess of it," has /not/ been met.
- Edsger Dijkstra, 1930-2002
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2002-10-31 0:07 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-09-26 14:51 Exhaustive simulator testing Zack Weinberg
2002-09-27 3:11 ` Richard Earnshaw
2002-09-27 10:16 ` Tom Tromey
2002-09-28 7:16 ` Richard Earnshaw
2002-10-02 7:29 ` Jeff Law
2002-10-02 14:00 ` Alexandre Oliva
2002-10-18 6:53 ` Ben Elliston
2002-10-30 14:21 ` Zack Weinberg
2002-10-31 8:54 ` Phil Edwards
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).