Re: [PATCH 1/2] daemon: use timeout for uninterruptible poll

From: Junio C Hamano <gitster@pobox.com>
Date: Apr 18, 2018, 21:51 UTC
Message-ID: <xmqqy3hkfais.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <nycvar.QRO.7.76.6.1804182251070.4241@ZVAVAG-6OXH6DA.rhebcr.pbec.zvpebfbsg.pbz>

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

Show 8 quoted lines

> Unless I am misunderstanding violently what you say, that is, in which
> case I would like to ask for a clarification why this patch (which does
> not change a thing unless NO_POLL is defined!) must be rejected, and while
> at it, I would like to ask you how introducing a layer of indirection with
> a full new function that is at least moderately misleading (as it would be
> named xpoll() despite your desire that it should do things that poll()
> does *not* do) would be preferable to this here patch that changes but a
> few lines to introduce a regular heartbeat check for platforms that

Our xwrite() and other xfoo() are to "fix" undesirable aspect of the underlying pure POSIX API to make it more suitable for our codebase. When pure POSIX poll() that requires the implementing or emulating platform pays attention to the children being waited on is not appropriate for the codepath we are using (i.e. the place where the patch is touching), it would be in line to introduce a "fixed" API that allows us to pass that information, so that we can build on top of that abstraction that is *not* pure POSIX abstraction, no? After all, you are the one who constantly whine that Git is implemented on POSIX API and it is inconvenient for other platforms.