Re: Fixing security vulnerabilities before new releases

Eshel Yaron <[email protected]> Wed, 29 Jul 2026 19:37:41 +0200
Newsgroups gmane.emacs.devel
Message-ID <[email protected]>
Hi,

Manuel Giraud <[email protected]> writes:

> Eshel Yaron <[email protected]> writes:
>
>> Sean Whitton <[email protected]> writes:
>>
>>> Eshel Yaron [29/Jul  1:16pm +02] wrote:
>>>>
>>>> Hmm...  There's an arbitrary code execution vulnerability that makes
>>>> opening files with Emacs unsafe.  Indeed, it affects existing Emacs
>>>> versions as well---that's a good reason to mitigate it ASAP and maybe
>>>> provide a security fix for Emacs 30, but it wouldn't make sense to
>>>> release another vulnerable version.
>>>> ISTM that if anything is a release-blocker, this is it.
>>>
>>> I don't understand your reasoning at all here.  We do not make the
>>> situation worse by releasing Emacs 31.  Therefore the issues are not
>>> interdependent.
>>
>> OK, let me try to clarify it further:
>> Releasing a new version invites more users to install and use it,
>> and encouraging users to run vulnerable software is at best a disservice.
>> True, it wouldn't make things worse for users who are already running a
>> vulnerable version, but I don't think that's the right bar to aim for.
>> Instead, we should make sure we make things better for them, by providing
>> a version that mitigates this issue ASAP.  AFAICT, releasing a vulnerable
>> Emacs 31.1 would delay that mitigation unnecessarily.
>>
>> To be clear, I think a mitigation can be rather simple/straightforward,
>> and it doesn't have to be the permanent solution.  I'd recommend
>> releasing Emacs 31 with a "bandaid" solution (I can suggest something if
>> you agree), and fixing it properly in a later release.
>
> I'm with you and Óscar on this one but FWIW I was not able to reproduce
> the code execution example given by Stefan.  Writing the following in
> /tmp/test.el and opening it doesn't do anything for me (I'm on current
> master):

[...]

> What am I missing?

I can readily reproduce it here, so your guess is as good as mine...
Did you try with emacs -Q?  Perhaps it's something in your config,
like a non-default value of vc-handled-backends, or maybe you have
trust-manager-mode enabled? :)


Best,

Eshel