Re: [PATCH] test-progress: fix test failures on big-endian systems

From: Junio C Hamano <gitster@pobox.com>
Date: Oct 23, 2019, 01:58 UTC
Message-ID: <xmqqpniojk8z.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <20191021184954.GA2526@sigill.intra.peff.net>

Jeff King <peff@peff.net> writes:

> But here's where it gets tricky. In addition to catching any size
> mismatches, this will also catch signedness problems. I.e., if we make
> OPT_INTEGER() use "intp", then everybody passing in &unsigned_var now
> gets a compiler warning. Which maybe is a good thing, I dunno.

Hmph, true.  I'd agree with back-burnering it for now.

Perhaps we'd fix the signedness issue one by one in a preparatory series before converting the value field to a union, if we want to pursue this idea further (in which I am mildly interested, by the way), but it does sound like it should be given lower priority.

Show 5 quoted lines

> So that's where I gave up. Converting between signed and unsigned
> variables needs to be done very carefully, as there are often subtle
> impacts (e.g., loop terminations). And because we have so many sign
> issues already, compiling with "-Wsign-compare", etc, isn't likely to
> help.

True.

Thanks.