[PATCH v2] Makefile: Fix elf/tst-trace tests
Adhemerval Zanella Netto
adhemerval.zanella@linaro.org
Mon Nov 2 19:05:00 GMT 2026
On 02/11/26 10:48, Yury Khrustalev wrote:
> The elf/tst-trace tests used
>
> ${ ... } $$@; $$(evaluate-test)
>
> that ignored any errors inside ${ ... }. These braces seem to be
> redundant and only mask any issues in the enclosed command.
>
> Also fix underlying issue in tst-ld-trace.py that parsed command
> line arguments incorrectly.
>
> Suggested-by: Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>
This patch looks ok, two minor points below.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
> ---
> elf/Makefile | 7 ++++---
> scripts/tst-ld-trace.py | 24 ++++++------------------
> 2 files changed, 10 insertions(+), 21 deletions(-)
>
> diff --git a/elf/Makefile b/elf/Makefile
> index 2c74e11946..058c5957b5 100644
> --- a/elf/Makefile
> +++ b/elf/Makefile
> @@ -3478,11 +3478,12 @@ $(objpfx)tst-trace$(1).out: $(objpfx)libtracemod1/libtracemod1.so \
> $(objpfx)libtracemod5/libtracemod5.so \
> $(..)scripts/tst-ld-trace.py \
> tst-trace$(1).exp
> - ${ $(PYTHON) $(..)scripts/tst-ld-trace.py \
> - "$(test-wrapper-env) $(elf-objpfx)$(rtld-installed-name) \
> + $(PYTHON) $(..)scripts/tst-ld-trace.py \
> + "$(test-wrapper-env) LD_TRACE_LOADED_OBJECTS=1 \
> + $(elf-objpfx)$(rtld-installed-name) \
> --library-path $(common-objpfx):$(strip $(2)) \
> $(objpfx)libtracemod1/libtracemod1.so" tst-trace$(1).exp \
> - } > $$@; $$(evaluate-test)
> + > $$@; $$(evaluate-test)
> endef
>
> $(eval $(call tst-trace-skeleton,1,))
> diff --git a/scripts/tst-ld-trace.py b/scripts/tst-ld-trace.py
> index 92c5b35bfc..d621207b79 100755
> --- a/scripts/tst-ld-trace.py
> +++ b/scripts/tst-ld-trace.py
> @@ -22,6 +22,7 @@ import argparse
> import os
> import subprocess
> import sys
> +import shlex
>
> try:
> subprocess.run
> @@ -54,10 +55,8 @@ def is_vdso(lib):
>
>
> def parse_trace(cmd, fref):
> - new_env = os.environ.copy()
> - new_env['LD_TRACE_LOADED_OBJECTS'] = '1'
> - trace_out = subprocess.run(cmd, stdout=subprocess.PIPE, check=True,
> - universal_newlines=True, env=new_env).stdout
> + trace_out = subprocess.run(shlex.split(cmd), stdout=subprocess.PIPE, check=True,
> + universal_newlines=True).stdout
A test-wrapper="WRAPPER_DEBUG=1 /path/wrap" now fails with
FileNotFoundError: 'WRAPPER_DEBUG=1', while the same string with shell=True
passes.
The manual/install.texi only documents 'cross-test-ssh.sh HOST', so this
isn't a blocker. But other tests uses shell=True (like
elf/tst-rtld-list-diagnostics.py).
> trace = []
> for line in trace_out.splitlines():
> line = line.strip()
> @@ -87,21 +86,10 @@ def parse_trace(cmd, fref):
> sys.exit(ret)
>
>
> -def get_parser():
> - parser = argparse.ArgumentParser(description=__doc__)
> - parser.add_argument('command',
> - help='command to run')
> - parser.add_argument('reference',
> - help='reference file to compare')
> - return parser
> -
> -
> def main(argv):
> - parser = get_parser()
> - opts = parser.parse_args(argv)
> - with open(opts.reference, 'r') as fref:
> - # Remove the initial 'env' command.
> - parse_trace(opts.command.split()[1:], fref)
> + command, reference = argv
> + with open(reference, 'r') as fref:
> + parse_trace(command, fref)
The 'import argparse' is now unused, I think it should be removed. A
wrong argument count or invalid option now ends in a 'ValueError', but
it should be ok (make check will fail anyway).
>
>
> if __name__ == '__main__':
More information about the Libc-alpha
mailing list