[RESEND PATCH 0/1] sim/testsuite: Trim extra path from arch
Tsukasa OI
research_trasio@irq.a4lg.com
Sat Aug 27 01:53:01 GMT 2022
Hello,
The patch is a RESEND of
<https://sourceware.org/pipermail/binutils/2022-July/121814.html>
but to the right mailing list and completely new cover letter.
Because my copyright assignment is complete and I know the right place to
talk, let me describe the details.
I first reproduced this bug while I'm adding simulator tests to RISC-V
but talking with AArch64 (with some real tests) would be better.
[How to reproduce]
1. Configure Binutils with aarch64-unknown-elf and build it
/src/binutils/configure \
--target=aarch64-unknown-elf \
--prefix=/opt/cross/aarch64-unknown-elf \
--enable-multilib
&& make
2. Run `make check-sim' and confirmed that the simulator tests "pass"
3. Intentionally try to fail the test by modifying
`sim/testsuite/aarch64/pass.s'
(replace the last line from "pass" to "fail")
4. Run `make check-sim' and "confirmed" that the simulator tests "pass"
[Analysis]
... Yes, something is going wrong. After the test, you can see the test
log in `sim/testsuite/aarch64/allinsn/testrun.log':
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Test run by builduser on Sat Aug 27 01:04:31 2022
Target is aarch64-unknown-elf
Host is x86_64-pc-linux-gnu
=== tests ===
Schedule of variations:
unix
Running target unix
Using /usr/share/dejagnu/baseboards/unix.exp as board description file for target.
Using /usr/share/dejagnu/config/unix.exp as generic interface file for target.
Using /src/binutils/sim/testsuite/config/default.exp as tool-and-target-specific interface file.
WARNING: Assuming target board is the local machine (which is probably wrong).
You may need to set your DEJAGNU environment variable.
Running /src/binutils/sim/testsuite/./aarch64/allinsn.exp ...
Can't find a compatible assembler
Executing on host: false /src/binutils/sim/testsuite/lib/compilercheck.c -E -o testsuite/./aarch64/allinsn/compilercheck.x (timeout = 300)
spawn -ignore SIGHUP false /src/binutils/sim/testsuite/lib/compilercheck.c -E -o testsuite/./aarch64/allinsn/compilercheck.x
compiler exited with status 1
Can't find a compatible C compiler
Testing adds.s on machine aarch64.
UNTESTED: ./aarch64/adds.s
[...]
Testing pass.s on machine aarch64.
UNTESTED: ./aarch64/pass.s
[...]
Testing xtn.s on machine aarch64.
UNTESTED: ./aarch64/xtn.s
testcase /src/binutils/sim/testsuite/./aarch64/allinsn.exp completed in 0 seconds
=== Summary ===
# of untested testcases 27
runtest completed at Sat Aug 27 01:04:31 2022
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
We clearly find that the testcase didn't find recently built assembler.
That assembler (and the linker) is supposed to be used because the
simulator itself (sim/Makefile) sets its configuration.
Quoting `sim/site.exp' and `sim/site-sim-config.exp':
set AS_FOR_TARGET_AARCH64 "/build/binutils/./gas/as-new"
set LD_FOR_TARGET_AARCH64 "/build/binutils/./ld/ld-new"
At last, we find `sim/testsuite/lib/sim-defs.exp'.
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
proc sim_init_toolchain {} {
[...]
set arch [sim_arch]
set ARCH [string map {- _} [string toupper $arch]]
foreach var {AS LD CC} {
set var_for_target "${var}_FOR_TARGET"
global $var_for_target
set var_for_target_arch "${var_for_target}_${ARCH}"
global $var_for_target_arch
if [info exists $var_for_target_arch] {
set $var_for_target [set $var_for_target_arch]
} else {
set $var_for_target ""
}
[...]
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
In here, it extracts {AS,LD,CC}_FOR_TARGET_AARCH64 and sets proper
{AS,LD,CC}_FOR_TARGET. At least, it is supposed to do so.
However, this block doesn't work because the `arch' variable returned by
the `sim_arch' function is "./aarch64". That is supposed to be "aarch64".
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
proc sim_arch {} {
global subdir
set arch "$subdir"
while { [file dirname $arch] != "." } {
set arch [file dirname $arch]
}
return "$arch"
}
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
`sim_arch' function is simple. Until the dirname part of the "arch" is not
"." (current directory), it trims the filename part.
For instance, if subdir is "A/B/C/D", this function returns "A". However,
if subdir is "./A/B/C/D", this function returns "./A", not "A".
In fact, actual subdir value here is "./aarch64" and we will get
"./aarch64" for [sim_arch]. As a result, it fails to extract proper
assembler/linker/compiler for given target.
[Fix: Patch Details]
After this patch, `sim/testsuite/aarch64/allinsn/testrun.log' will
look like this:
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Test run by builduser on Sat Aug 27 01:34:29 2022
Target is aarch64-unknown-elf
Host is x86_64-pc-linux-gnu
=== tests ===
Schedule of variations:
unix
Running target unix
Using /usr/share/dejagnu/baseboards/unix.exp as board description file for target.
Using /usr/share/dejagnu/config/unix.exp as generic interface file for target.
Using /src/binutils/sim/testsuite/config/default.exp as tool-and-target-specific interface file.
WARNING: Assuming target board is the local machine (which is probably wrong).
You may need to set your DEJAGNU environment variable.
Running /src/binutils/sim/testsuite/./aarch64/allinsn.exp ...
Found a compatible assembler
Can't execute C compiler
Testing adds.s on machine aarch64.
Executing on host: /build/binutils/./gas/as-new /src/binutils/sim/testsuite/./aarch64/adds.s -I/src/binutils/sim/testsuite/./aarch64 -o testsuite/./aarch64/allinsn/adds.s.o (timeout = 300)
spawn -ignore SIGHUP /build/binutils/./gas/as-new /src/binutils/sim/testsuite/./aarch64/adds.s -I/src/binutils/sim/testsuite/./aarch64 -o testsuite/./aarch64/allinsn/adds.s.o
Executing on host: /build/binutils/./ld/ld-new testsuite/./aarch64/allinsn/adds.s.o -o testsuite/./aarch64/allinsn/adds.s.x (timeout = 300)
spawn -ignore SIGHUP /build/binutils/./ld/ld-new testsuite/./aarch64/allinsn/adds.s.o -o testsuite/./aarch64/allinsn/adds.s.x
./aarch64/run testsuite/./aarch64/allinsn/adds.s.x
spawn ./aarch64/run testsuite/./aarch64/allinsn/adds.s.x
pass
PASS: aarch64 adds.s
[...]
Testing pass.s on machine aarch64.
Executing on host: /build/binutils/./gas/as-new /src/binutils/sim/testsuite/./aarch64/pass.s -I/src/binutils/sim/testsuite/./aarch64 -o testsuite/./aarch64/allinsn/pass.s.o (timeout = 300)
spawn -ignore SIGHUP /build/binutils/./gas/as-new /src/binutils/sim/testsuite/./aarch64/pass.s -I/src/binutils/sim/testsuite/./aarch64 -o testsuite/./aarch64/allinsn/pass.s.o
Executing on host: /build/binutils/./ld/ld-new testsuite/./aarch64/allinsn/pass.s.o -o testsuite/./aarch64/allinsn/pass.s.x (timeout = 300)
spawn -ignore SIGHUP /build/binutils/./ld/ld-new testsuite/./aarch64/allinsn/pass.s.o -o testsuite/./aarch64/allinsn/pass.s.x
./aarch64/run testsuite/./aarch64/allinsn/pass.s.x
spawn ./aarch64/run testsuite/./aarch64/allinsn/pass.s.x
fail
status: 0
output: fail
pattern: pass
FAIL: aarch64 pass.s (execution)
[...]
Testing xtn.s on machine aarch64.
Executing on host: /build/binutils/./gas/as-new /src/binutils/sim/testsuite/./aarch64/xtn.s -I/src/binutils/sim/testsuite/./aarch64 -o testsuite/./aarch64/allinsn/xtn.s.o (timeout = 300)
spawn -ignore SIGHUP /build/binutils/./gas/as-new /src/binutils/sim/testsuite/./aarch64/xtn.s -I/src/binutils/sim/testsuite/./aarch64 -o testsuite/./aarch64/allinsn/xtn.s.o
Executing on host: /build/binutils/./ld/ld-new testsuite/./aarch64/allinsn/xtn.s.o -o testsuite/./aarch64/allinsn/xtn.s.x (timeout = 300)
spawn -ignore SIGHUP /build/binutils/./ld/ld-new testsuite/./aarch64/allinsn/xtn.s.o -o testsuite/./aarch64/allinsn/xtn.s.x
./aarch64/run testsuite/./aarch64/allinsn/xtn.s.x
spawn ./aarch64/run testsuite/./aarch64/allinsn/xtn.s.x
pass
PASS: aarch64 xtn.s
testcase /src/binutils/sim/testsuite/./aarch64/allinsn.exp completed in 0 seconds
=== Summary ===
# of expected passes 26
# of unexpected failures 1
runtest completed at Sat Aug 27 01:34:29 2022
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
OK, we successfully failed. If we restore `sim/testsuite/aarch64/pass.s',
it completely succeeds.
This patch trims "./" part of "./arch" using the `file tail' function.
Even if the subdir input is "aarch64", it returns "aarch64".
With this patch, we can now successfully test the simulator.
This is effectively a prerequisite of my Zmmul patchset on RISC-V because
I (and another person who also submitted Zmmul patchset) accidentally broke
the simulator while adding Zmmul and my Zmmul patchset contains a test case
to make sure that the simulator works as expected.
Thanks,
Tsukasa
Tsukasa OI (1):
sim/testsuite: Trim extra path from arch
sim/testsuite/lib/sim-defs.exp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
base-commit: 46e59b72f21029f2a863e3828cec76a03283b49d
--
2.34.1
More information about the Gdb-patches
mailing list