Re: [PATCH 1/4] GDBSERVER: Listen on a unix domain (instead of TCP) socket if requested.

[John Darrington <john at darrington dot wattle dot id dot au>]

On Mon, Oct 29, 2018 at 03:51:55PM +0000, Simon Marchi wrote:
     On 2018-10-29 5:11 a.m., Rainer Orth wrote:
     > Hi John,
     > 
     >> However I've checked in a fix for this issue, and tested it by building
     >> natively with a hacked set of standard include headers.
     > 
     > you always need to post patches here, if only for reference.
     
Doesn't that make the gdb-cvs list completely redundant?

     Not only should you post here the patches you push as obvious, but I don't
     think that this:
     
     https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=98a17ece013cb94cd602496b9efb92b8816b3953
     
     falls under the obvious rule:

But a number of people had complained that their build was broken, and
this was a fix for that.   I judged that in consideration of those
people fixding their problem was more important than strict observance 
of protocol.
     
     I can't judge whether the patch is right or not with a quick glance, but it
     certainly is complex enough to warrant a discussion (as Rainer's reply below
     shows).

     
     Additionally, it seems like the initial 4-patch series was pushed without
     explicit approval from a maintainer (at least I can't find any).  Next time,
     please wait to have an approval before pushing.  If you are not sure whether
     a reply constitutes an a approval, it's better to ask the maintainer to
     clarify.

All of those patches were certainly discussed.   In the past, when I've
followed up a person who has commented on a patch, but been vague about
approval, I have had either a piqued response; or no response at all.

If you think it necesary however I can revert anything you think hasn't
had enough discussion.

     
     > Besides, we're currently very inconsistent here (haven't checked which
     > part of that is due to your code): most places use AF_UNIX, only two use
     > AF_LOCAL instead (common/netstuff.c, gdbserver/remote-utils.c), and your
     > configure check only checks for AF_LOCAL.  I believe we should
     > canonicalize for one of the two and allow for systems that define only
     > one or the other.
     
     mingw defines AF_UNIX, so I would tend to go towards that route.  Any of you
     knows what happens at runtime when you try to bind a AF_UNIX socket on mingw/Windows?
     
AS I understand it, it depends upon which version of windows.
Apparently recent versions have local ssockets, but older ones do not.

See https://blogs.msdn.microsoft.com/commandline/2017/12/19/af_unix-comes-to-windows/

J'