Re: [PATCH] testsuite: support mold linker

["Martin Liška" <mliska@suse.cz>]

On 12/5/22 13:22, Jan Beulich wrote:
> On 05.12.2022 13:10, Martin Liška wrote:
>> --- a/binutils/testsuite/binutils-all/objcopy.exp
>> +++ b/binutils/testsuite/binutils-all/objcopy.exp
>> @@ -652,8 +652,8 @@ proc strip_test_with_saving_a_symbol { } {
>>  
>>      set exec_output [binutils_run $NM "$NMFLAGS $objfile"]
>>      set exec_output [prune_warnings $exec_output]
>> -    if {![regexp {^([0-9a-fA-F]+)?[ ]+[TD] main} $exec_output] \
>> -         && ![regexp {^([0-9a-fA-F]+)?[ ]+T _main} $exec_output]} {
>> +    if {![regexp {^([0-9a-fA-F]+)?[ ]+[tTD] main} $exec_output] \
> 
> Further below, for libbacktrace, you also include 'd' in the check (and 't'
> was already there). Could you clarify why 't' alone isn't sufficient there,
> but is sufficient here?

Yeah, it is sufficient.

> 
>> +         && ![regexp {^([0-9a-fA-F]+)?[ ]+[tT] _main} $exec_output]} {
> 
> While orthogonal to the purpose of the patch I still wonder whether you
> wouldn't better take the opportunity and fold the two regexp-s (also
> elsewhere):
> 
>     if {![regexp {^[0-9a-fA-F]*[ ]+[tTD] _?main} $exec_output]} {

Yep, works for me.

> 
> (with a simplification to the earlier part also included). Looking at
> strip_executable_with_saving_a_symbol, which you also adjust, the [TtD]
> vs [tT] difference looks to be accidental rather than deliberate.

Correct, D is really accidental value here.

Lemme send V2.

Martin

> 
> Jan