Re: Understanding why getaddrinfo_a and __gai_enqueue_request use recursive locks

[Will Hawkins <whh8b at virginia dot edu>]

On Wed, Sep 27, 2017 at 1:53 AM, Carlos O'Donell <carlos@redhat.com> wrote:
> On 09/26/2017 12:39 PM, Will Hawkins wrote:
>> On Tue, Sep 26, 2017 at 2:06 AM, Will Hawkins <whh8b@virginia.edu> wrote:
>>> On Tue, Sep 26, 2017 at 12:42 AM, Carlos O'Donell <carlos@redhat.com> wrote:
>>>> Can you test removing them and ga_test.c?
>>>>
>>>> diff --git a/resolv/Makefile b/resolv/Makefile
>>>> index ec7e4fd..8a822e0 100644
>>>> --- a/resolv/Makefile
>>>> +++ b/resolv/Makefile
>>>> @@ -101,10 +101,6 @@ routines                += $(libnss_dns-routines) $(libresolv-routines)
>>>>  static-only-routines    += $(libnss_dns-routines) $(libresolv-routines)
>>>>  endif
>>>>
>>>> -ifeq (yesyes,$(build-shared)$(have-thread-library))
>>>> -tests: $(objpfx)ga_test
>>>> -endif
>>>> -
>>>>  ifeq ($(run-built-tests),yes)
>>>>  ifneq (no,$(PERL))
>>>>  tests-special += $(objpfx)mtrace-tst-leaks.out
>>>> @@ -134,8 +130,6 @@ $(objpfx)libnss_dns.so: $(objpfx)libresolv.so
>>>>  # The asynchronous name lookup code needs the thread library.
>>>>  $(objpfx)libanl.so: $(shared-thread-library)
>>>>
>>>> -$(objpfx)ga_test: $(objpfx)libanl.so $(shared-thread-library)
>>>> -
>>>>  $(objpfx)tst-res_hconf_reorder: $(libdl) $(shared-thread-library)
>>>>  tst-res_hconf_reorder-ENV = RESOLV_REORDER=on
>>>>
>>>> ---
>>>>
>>>
>>> I am testing this as we speak but don't think that I will know the
>>> answer before I head to sleep tonight. I will respond as soon as my
>>> tests have completed their execution.
>>
>> Testing of this 'patch' is complete. I applied the change then used
>> the instructions on
>> https://sourceware.org/glibc/wiki/Testing/Testsuite#Testing_just_one_test
>> to test. Here's how I did it:
>>
>> From the build directory, I did
>> rm -f resolv/*.out
>> make subdirs=resolv check
>>
>> and everything ran the way I expected (the same number of test
>> failures and passes).
>>
>> Thanks again for all your help!
> Awesome, thanks. I'll see about posting a patch to do that, and include
> you in the testing notes. Alternatively you can post the patch yourself
> and I'll approve it, the changes are mechanical and below the copyright
> limit, but any future work would require copyright assignment :-)

If you aren't in any hurry to have this added, then I'd like to do it.
It will give me practice at the submission process while I work on the
other patch for removing the locks held by the callee function. Just
let me know whether it would be okay to wait for a few days on this
patch. If so, I'll take a crack at it! I would certainly not mind you
moving ahead if you wanted to move quicker. I just thought I'd offer
to do it for the experience and to keep you from having to use your
valuable time!

Thanks again for all the help!
Will

>
> --
> Cheers,
> Carlos.