Re: [PATCH v5 1/8] Move malloc hooks into a compat DSO

[Siddhesh Poyarekar <siddhesh@sourceware.org>]

On 7/6/21 7:59 PM, Florian Weimer via Libc-alpha wrote:
> * Siddhesh Poyarekar:
> 
>>> In general, it looks like it's time to add support for mtrace tests to
>>> the setsuite (that is, tests-mtrace like tests-time64).
>>
>> ... or tests-mcheck/tests-malloc-check.  Would you be OK with having
>> the test driver call mtrace() at all times?
> 
> Could we enable mtrace directly via LD_PRELOAD (or static linking),
> without calling the function?
> 
> There's also the matter of not flagging potentially leaking allocations
> from the test harness itself.

We could link tests with an object that calls mtrace through the 
initialization hook like we do for tests-mcheck, but it will be 
equivalent to calling mtrace() in the driver.  If anything, it may end 
up getting called earlier, possibly recording dynamic linker allocations.

>> I forgot to ask, would you like that to be part of this patchset or
>> would you be OK with the testing enhancement coming in later?  Mtrace
>> already has some testing coverage through tst-mtrace for example and
>> other tests using mtrace.
> 
> I don't have a strong preference.  Given the open questions, it's
> probably best to delay the cleanup.

OK thanks.  I won't delay too long, just that I'd like it if it doesn't 
block this patchset.

Thanks,
Siddhesh