severity 367058 important thanks On Fri, Jun 02, 2006 at 01:08:59AM +0200, Adeodato Simó wrote: > > Either way, there is still a bug in gnupg-agent here; at most, the > > x11-common workaround affects the severity of the gnupg-agent bug, but it's > > still a gnupg-agent bug. And it's still a bug that would manifest when > > installing gnupg-agent on a sarge system (partial upgrades), so it should > > still be fixed.
> I understand this reasoning. However, this is my take on the matter: the > current script is well written, and makes reasonably sure that if the > final exit status is unsuccessful, it's not due to something under its > control, but outside (e.g., a badly-written gpg-agent.conf). When this > occurs, the script just ends with a non-zero status, as it should. > Since it's a Xsession script, the author may think of what should happen > if an error occurs: > (1) just leave the caller decide (current situation) > (2) ensure an error won't cause the X11 startup process to abort, by > adding "|| true" to the `gpg-agent --daemon` invocation > (3) ensure an error will cause the X11 startup process to abort, by > adding "|| exit 1" to the `gpg-agent --daemon` invocation > And, in my opinion, no Xsession script should get away from (1), so in > my opinion the script is fine as it is, and if I were the gnupg-agent > maintainer, this'd be wishlist+wontfix, since I can't think of (2) nor > (3) as the fix you ask for above. > I merged because I took this bug not as particular to gpg-agent, but to > any X11 init script: I saw here an user saying that user errors that pop > up in such scripts should not finish the session, and since this was > resolved elsewhere in the distribution, I pointed him there. > > Well, I don't agree that the change to x11-common was a correct fix; we have > > every reason to demand the same high quality of error handling for Xsession > > scripts as we do for other scripts in Debian, which is undermined by this > > casual use of set +e as a workaround. Isn't it reasonable to expect that I > > may *want* a failed Xsession script to cause the session to abort? > Yeah, we may, though as a distribution default, I think not aborting the > session is a better default. And then, is a non-zero exit status from > gpg-agent a failed Xsession script or not? I think it is, and the script > already ends with the appropriate exit status, letting the parent decide > what to do. > But alas, perhaps there's a number (4) option I did not see. For the > moment, I can't think how this bug can remain to be RC. On Thu, Jun 01, 2006 at 09:51:19PM -0400, David Nusinow wrote: > > Well, I don't agree that the change to x11-common was a correct fix; we have > > every reason to demand the same high quality of error handling for Xsession > > scripts as we do for other scripts in Debian, which is undermined by this > > casual use of set +e as a workaround. Isn't it reasonable to expect that I > > may *want* a failed Xsession script to cause the session to abort? > You may, but I think you're in the minority. I don't think a user who > installs some rogue package with a broken script deserves to have the > ability to log in to X ripped out from under them. It's the maintainer's > job to ensure that the script works, and the user shouldn't have to suffer > unnecessarily for mistakes. While I'm all for forcing devices to ensure > quality, bringing down all of X because gpg-agent doesn't start is pretty > extreme. > > Either way, there is still a bug in gnupg-agent here; at most, the > > x11-common workaround affects the severity of the gnupg-agent bug, but it's > > still a gnupg-agent bug. And it's still a bug that would manifest when > > installing gnupg-agent on a sarge system (partial upgrades), so it should > > still be fixed. > Indeed. If the change that I made is causing maintainers to divest > themselves of responsibility for their packages, then I'll revert it, no > questions. The goal is to protect users, not maintainers. You both make fair points, that there is room for disagreement here about whose responsibility it is to make sure Xsession scripts don't fail out and abort the X session. It is still my opinion that this is a bug in gnupg-agent, because the default behavior desired by most users, AFAICT, is that installing gnupg-agent will give a gpg-agent to those users with a correctly configured env while not interfering with the X sessions of those without one. So I think the default behavior here should be for gnupg-agent to catch its own errors here and return success, independent of how any other Xsession hooks might operate. But I don't think this is RC, so downgrading to important. Cheers, -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. [EMAIL PROTECTED] http://www.debian.org/
signature.asc
Description: Digital signature
