[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
bug#31497: 27.0.50; vc-hg doesn't separate standard error from standard
From: |
Lars Ingebrigtsen |
Subject: |
bug#31497: 27.0.50; vc-hg doesn't separate standard error from standard output |
Date: |
Wed, 09 Oct 2019 23:57:15 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/27.0.50 (gnu/linux) |
Philipp Stephani <p.stephani2@gmail.com> writes:
> Create a Mercurial repository owned by root and add a hgrc file, e.g.:
>
> $ mkdir -p /tmp/hg
> $ cd /tmp/hg
> $ sudo -s
> # hg init
> # echo a > a.txt
> # hg addremove
> # hg commit -m 'commit message'
> # touch .hg/hgrc
>
> Then visit a.txt in Emacs as non-root:
>
> $ emacs -Q a.txt
>
> Instead of the revision number, the status line will show
> "Hg-not trusting file /tmp/hg/.hg/hgrc ..."
>
> This is because the status line runs something like
>
> $ hg log -l 1 -r . -T '{rev}'
>
> but that outputs
>
> not trusting file /tmp/hg/.hg/hgrc from untrusted user root, group root
> not trusting file /tmp/hg/.hg/hgrc from untrusted user root, group root
>
> on stderr, and Emacs merges stderr and stdout. Rather it should ignore
> stderr in this case.
I can confirm that this bug is still present in Emacs 27. The command
is started from vc-doc-command, which uses start-file-process, which
uses start-process, which doesn't even allow separating stderr from
stdin.
(In general, Emacs stderr/stdin handling is very frustrating.)
Would it make sense to alter this function
(defun vc-hg--run-log (template rev path)
(ignore-errors
(with-output-to-string
(if path
(vc-hg-command
standard-output 0 nil
"log" "-f" "-l1" "--template" template path)
(vc-hg-command
standard-output 0 nil
"log" "-r" rev "-l1" "--template" template)))))
to filter out the "not trusting" lines, or is this just one of a
gazillion error messages that can be output here?
not trusting file /tmp/hg/.hg/hgrc from untrusted user root, group root
not trusting file /tmp/hg/.hg/hgrc from untrusted user root, group root
4d872e
I though perhaps we could redirect the error output with "-E", but "hg
log" doesn't seem to support that.
--
(domestic pets only, the antidote for overdose, milk.)
bloggy blog: http://lars.ingebrigtsen.no
- bug#31497: 27.0.50; vc-hg doesn't separate standard error from standard output,
Lars Ingebrigtsen <=