[v4] Add helper script for glibc debugging

  Hi Gabriel,

Thanks for writing this. The patch looks good to me.

I have one small suggestion. What do you think about this?:


The idea being that this will allow invoking (non-testcase) programs with
arguments, like:

./debugglibc.sh -i iconv/iconv_prog -f ASCII -t UTF-8 ascii.txt

I don't necessarily want this patch to go through more versions. It already
adds value and I think it's ready to go in today. Just wanted to enable
another usecase. Perhaps we can discuss it once the current version is in.

Cheers,
Arjun
  

Hi, Arjun,

On Mon, 30 Sep 2019, Arjun Shankar wrote:

>The idea being that this will allow invoking (non-testcase) programs with
>arguments, like:

I haven't thought about this use case before, but it sounds good to me. :)

>./debugglibc.sh -i iconv/iconv_prog -f ASCII -t UTF-8 ascii.txt

I think we might need some sort of coordination between the optional
arguments to debugglibc.sh and these arguments to the program being tested.
Otherwise, the position of arguments such as "-e ENVVAR=VALUE" might break
this new use case.

For instance, currently, it doesn't matter if the optional arguments are
passed before or after the path to the test case, so, both:

./debugglibc.sh -b pthread_join nptl/tst-exec1
and
./debugglibc.sh nptl/tst-exec1 -b pthread_join

work.

>       TESTCASE=$$1
>+      COMMANDLINE="$$@"
>+      break
>       ;;

With this suggested change, that would no longer work, which is not
necessarily a problem, but caught my attention.

>+    --)
>+      shift
>+      TESTCASE=$$1
>+      COMMANDLINE="$$@"
>+      break
>+      ;;
>     *)

This option, on the other hand, has no conflicts with the current
behavior, but it would need some documentation in the --help message.

>I don't necessarily want this patch to go through more versions. It
>already adds value and I think it's ready to go in today. Just wanted to
>enable another usecase. Perhaps we can discuss it once the current
>version is in.

Since you already wrote the diff, would you also like to write the commit
message and send it as a patch?

Cheers,
Gabriel