Re: minor patch required to compile git 2.26.1 on Oracle Solaris 10 with Oracle Studio

From: Jeff King <peff@peff.net>
Date: Apr 22, 2020, 08:09 UTC
Message-ID: <20200422080950.GA464427@coredump.intra.peff.net>
In-Reply-To: <xmqqzhb8rqa6.fsf@gitster.c.googlers.com>

On Sat, Apr 18, 2020 at 10:35:45PM -0700, Junio C Hamano wrote:

Show 7 quoted lines

> Dennis Clarke <dclarke@blastwave.org> writes:
> 
> > Very minor patch required :
> 
> But both size_t and unsigned int are wrong types, no?  Shouldn't we
> be using socklen_t (and we seem to have autoconf support to figure
> out an appropriate socklen_t fallback typedef)?

That is generally the right type according to POSIX, but I think there's something even more subtle going on. If we're defining our own inet_ntop(), then the type shouldn't really matter, should it?

Dennis didn't show us the compiler errors, but my suspicion is that it is complaining because it is seeing the definition of inet_ntop() already from a system header, and ours does not match.

Which implies that NO_INET_NTOP should not be set in the first place. I think this is the same issue discussed in:

  https://lore.kernel.org/git/CAH8yC8m3JFvEcfFF3z1rrRnEPK-adHGObmkOhNZiph7QJKUWqA@mail.gmail.com/

with a patch (which needs at least a signoff added) in:

  https://lore.kernel.org/git/CAH8yC8kaWXbN+RYMJnM9em7KKW54+N07JtyS1MZk0qppD=m2BA@mail.gmail.com/

Dennis, does building with:

  make NO_INET_NTOP= NO_INET_PTON=

help?

-Peff