Re: [PATCH] gc: use temporary file for editing crontab

From: "brian m. carlson" <sandals@crustytoothpaste.net>
To: Derrick Stolee <derrickstolee@github.com>
Cc: "Johannes Schindelin" <Johannes.Schindelin@gmx.de>,
	git@vger.kernel.org, "Junio C Hamano" <gitster@pobox.com>,
	"Derrick Stolee" <stolee@gmail.com>,
	"Renato Botelho" <garga@freebsd.org>,
	"Todd Zullinger" <tmz@pobox.com>,
	"Đoàn Trần Công Danh" <congdanhqx@gmail.com>
Subject: Re: [PATCH] gc: use temporary file for editing crontab
Date: Tue, 23 Aug 2022 21:15:28 +0000	[thread overview]
Message-ID: <YwVDcO/V+zx2iy4I@tapette.crustytoothpaste.net> (raw)
In-Reply-To: <9e737b4b-4a17-09d5-6452-4ca5eef3d9da@github.com>

[-- Attachment #1: Type: text/plain, Size: 2031 bytes --]

On 2022-08-23 at 17:06:03, Derrick Stolee wrote:
> On 8/23/2022 5:12 AM, Johannes Schindelin wrote:
> > Hi brian,
> > 
> > On Tue, 23 Aug 2022, brian m. carlson wrote:
> 
> >> +	tmpedit = mks_tempfile_t(".git_cron_edit_tmpXXXXXX");
> >> +	if (!tmpedit)
> >> +		return error(_("failed to create crontab temporary file"));
> > 
> > It might make sense to use the same `goto out;` pattern here, to make it
> > easier to reason about the early exit even six years from now.
> > 
> > We do not even have to guard the `close_tempfile_gently()` behind an `if
> > (tempfile)` conditional because that function handles `NULL` parameters
> > gently.

I can do that.  I'll need to make sure we initialize the pointer to NULL
first.

> This is focused only on the cron integration, which is not used on Windows,
> so I'm not worried about that.

Correct.  The only place this could go wrong is Cygwin, but I believe it
has the proper behaviour (and if not, lots of stuff will be broken).

> I was initially worried that we lost the fclose(cron_in), but of course it
> is handled by the close_tempfile_gently() at the end.

Yup.  I originally called fclose here and glibc screamed at me about a
double-free, so the fclose definitely should be removed.  I'll mention
this in the commit message as well.

> Here's the crux of the matter: we are no longer using stdin but
> instead passing an argument to point to a file with our desired
> schedule. I tested that this worked on my machine, and I'm glad
> this use is the POSIX standard.
> 
> There is something wrong with this patch: it needs to update
> t/helper/test-crontab.c in order to pass t7900-maintenance.sh.

Will fix.

> While the memory release is nice, I also think it would be good to use
> delete_tempfile() so the temporary file is deleted within this method,
> not waiting until the end of the process to do that cleanup.

Sounds good.  I'll include that in a v2.
-- 
brian m. carlson (he/him or they/them)
Toronto, Ontario, CA

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 263 bytes --]