Re: [PATCH v3 3/3] trace.h: remove never-used TRACE_CONTEXT

From: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date: Feb 20, 2022, 12:38 UTC
Message-ID: <220220.86fsodd7sm.gmgdl@evledraar.gmail.com>
In-Reply-To: <xmqq5yp9wxt1.fsf@gitster.g>

On Sun, Feb 20 2022, Junio C Hamano wrote:

[I forgot to CC Karsten Blees]

Show 17 quoted lines

> Ævar Arnfjörð Bjarmason  <avarab@gmail.com> writes:
>
>> The definition of "TRACE_CONTEXT" as "__FILE__" added in
>> e05bed960d3 (trace: add 'file:line' to all trace output, 2014-07-12)
>> appeared between v6[1] and v7[2] of the series that added it.
>>
>> It does not appear to have been something anybody asked for, and
>> doesn't seem to have been used by anyone since then to override it to
>> something other than the default __FILE__.
>
> Sorry, but I do not quite follow.  How can we claim nobody used it,
> given that it should be usable with CFLAGS=-DTRACE_CONTEXT=ANYTHING?
>
> On the other hand, I do not see how it is hurting anybody to have
> this indirection today.
>
> Or am I missing something obvious?

I don't know that, but:

 * From the list discussion at the time it seems nobody asked for this,
   it was just something the author "snuck in" between re-rolls.

 * I didn't find any subsequent on-list discussion of people actually
   using this.

Both leads me to believe that this was a neat-at-the-time but never used for anything outside one-off testing feature.

Perhaps you'd like a v4 without this. It isn't strictly needed, but where I'm going with this series is improving the usage.c output/passing of these __{FILE,LINE,FUNCTION}__.

I don't need it now, but if we'd eventually like to unify trace & trace2 we'll need to deal with this part, since those functions won't deal well with replacing a __FILE__ "just pass whatever data you'd like here".