Re: [PATCH] [AArch64,v5] MTE corefile support

[Pedro Alves <pedro@palves.net>]

Hi!

This version looks good to me.  Just one nit in the testcase.  No need to repost for this.

On 2022-07-11 11:57 a.m., Luis Machado wrote:
> +foreach_with_prefix mode {"sync" "async"} {
> +
> +    if {$mode == "async"} {
> +	lappend compile_flags "additional_flags=-DASYNC"
> +    }
> +
> +    standard_testfile
> +    set executable "${testfile}-${mode}"
> +    if {[prepare_for_testing "failed to prepare" ${executable} ${srcfile} ${compile_flags}]} {
> +	return -1

If we hit this return while testing "sync", then "async" won't be tested either.  Similarly
for the other returns below.  To avoid scenarios like this (and in the future if the code evolves
to have more early returns), it's IMO better/safer to move the body of the foreach_with_prefix
to a procedure, like:

 proc test {mode} {
     global srcfile testfile
 
     if {$mode == "async"} {
 	lappend compile_flags "additional_flags=-DASYNC"
     }
 
     standard_testfile
     set executable "${testfile}-${mode}"
     if {[prepare_for_testing "failed to prepare" ${executable} ${srcfile} ${compile_flags}]} {
 	return -1
 
     ...
 }

 foreach_with_prefix mode {"sync" "async"} {
    test $mode
 }

Pedro Alves

> +    }
> +    set binfile [standard_output_file ${executable}]
> +
> +    if ![runto_main] {
> +	untested "could not run to main"
> +	return -1
> +    }
> +
> +    # Targets that don't support memory tagging should not execute the
> +    # runtime memory tagging tests.
> +    if {![supports_memtag]} {
> +	unsupported "memory tagging unsupported"
> +	return -1
> +    }