You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Yes, the current situation is fixed, but this is the second time we've received bug reports about out of date files. Do you have any method in place to prevent the problem from happening again in the future?
thanks for the ping, on it. hopefully we can turn key management over to the hackage foundation and professionalize all this in the next year or so. cc @emilypi
Reacted by Robert Hensing, Stuart Popejoy and Emily Pillmore
I am not a holder of the appropriate keys, but I'm reaching out to them now to try to resolve this. Additionally, now that the HF is more stabilized, I will work to try to get this and other keyholding business taken care of by the foundation as a standard place rather than having it distributed in ways that are a bit rather fiddly and can unfortunately fall through the cracks.
I wish I could help more, but I don't have the keys to perform the resigning. That said, I don't believe cabal install or any recent versions of stack are contingent on this mirrors file, so I am curious as a general proposition what workflows are breaking.
BTW, I agree giving operational keys to folks from the HF is a good idea. The root key holders will need to issue and sign an updated root.json that lists the operational keys (mirrors, snapshot, timestamp etc).
It might also be a good opportunity to add more root key holders & drop some who are no longer active.
While having more key holders would help in a pinch, the bigger problem is that no one is tracking this task on a calendar to do the resign before the expiration. Just like with SSL/TLS certs, someone needs this task on a calendar with reminders, and we need to do the resign before the expiration. Adding key holders is nice, but we should first figure out the process for how that task is tracked and who is responsible for the tracking. We should not be waiting until the expiration has passed.
EDIT: It's really frustrating that I'm repeating what was stated back in 2019 (almost 3 years ago!!):
Yes, the current situation is fixed, but this is the second time we've received bug reports about out of date files. Do you have any method in place to prevent the problem from happening again in the future?
Reacted by Duncan Coutts, Andrew Cowie and junji hashimoto
Reacted by Robert Hensing, Oleg Grenrus, Patrick Reader, Pete Murphy and junji hashimoto
Perhaps the expiry times could be made shorter: that way, people will be more used to doing it regularly, and might even think to partially automate it.
Ok, this is now resolved again for the time being. I agree the process is very bad, and that changes were never made to the procedures to keep this from reoccuring. I'm going to follow up immediately with the HF to try to put this into a better situation -- especially since we need a root resigning in approx the next 60 days as well.
I'm currently considering working around this by disabling the expiry checks by default in Stack's usage of hackage-security.
Since this issue is still open and it was not obvious to me, please note that the checks indeed have been disabled in newer Stack versions, by commit commercialhaskell/stack@4d54687 which was released in Stack 2.1.3.
If you come here because you're using Stack, you might see this error because you're using an old version.
Activity
gbaz commentedon Jul 1, 2019
Should be fixed now?
gbaz commentedon Jul 2, 2019
confirmed.
snoyberg commentedon Jul 2, 2019
Yes, the current situation is fixed, but this is the second time we've received bug reports about out of date files. Do you have any method in place to prevent the problem from happening again in the future?
roberth commentedon Jan 2, 2021
It has expired again. Current value:
gbaz commentedon Jan 2, 2021
thanks for the ping, on it. hopefully we can turn key management over to the hackage foundation and professionalize all this in the next year or so. cc @emilypi
gbaz commentedon Jan 2, 2021
ok, new one should be deployed thanks to duncan.
roberth commentedon Jan 2, 2021
I can confirm that it works. Thank you gbaz and duncan!
emilypi commentedon Jan 2, 2021
Cheers! Thanks @roberth and @gbaz
tp-woven commentedon Jun 2, 2022
I'm seeing this error again:
"expires":"2022-06-01T23:59:59Z"gbaz commentedon Jun 2, 2022
on it, thanks for the report.
gbaz commentedon Jun 2, 2022
I am not a holder of the appropriate keys, but I'm reaching out to them now to try to resolve this. Additionally, now that the HF is more stabilized, I will work to try to get this and other keyholding business taken care of by the foundation as a standard place rather than having it distributed in ways that are a bit rather fiddly and can unfortunately fall through the cracks.
jkoppel commentedon Jun 2, 2022
I am also running into this.
Also, today is the deadline for ICFP artifact evaluation, which makes this particularly dastardly.
gbaz commentedon Jun 2, 2022
I wish I could help more, but I don't have the keys to perform the resigning. That said, I don't believe cabal install or any recent versions of stack are contingent on this mirrors file, so I am curious as a general proposition what workflows are breaking.
4 remaining items
dcoutts commentedon Jun 2, 2022
I'll fix this now by re-signing the mirrors file. Will provide an update here when it's done.
But yes it would be wise to hand out a couple more operational keys. That'll need the root key holders to re-sign.
dcoutts commentedon Jun 2, 2022
Ok, next step is to upload it:
(It can be canonicalised or not, doesn't matter except file size)
dcoutts commentedon Jun 2, 2022
BTW, I agree giving operational keys to folks from the HF is a good idea. The root key holders will need to issue and sign an updated root.json that lists the operational keys (mirrors, snapshot, timestamp etc).
It might also be a good opportunity to add more root key holders & drop some who are no longer active.
ketzacoatl commentedon Jun 2, 2022
@dcoutts / @gbaz / etc
While having more key holders would help in a pinch, the bigger problem is that no one is tracking this task on a calendar to do the resign before the expiration. Just like with SSL/TLS certs, someone needs this task on a calendar with reminders, and we need to do the resign before the expiration. Adding key holders is nice, but we should first figure out the process for how that task is tracked and who is responsible for the tracking. We should not be waiting until the expiration has passed.
EDIT: It's really frustrating that I'm repeating what was stated back in 2019 (almost 3 years ago!!):
pxeger commentedon Jun 2, 2022
Perhaps the expiry times could be made shorter: that way, people will be more used to doing it regularly, and might even think to partially automate it.
gbaz commentedon Jun 2, 2022
Ok, this is now resolved again for the time being. I agree the process is very bad, and that changes were never made to the procedures to keep this from reoccuring. I'm going to follow up immediately with the HF to try to put this into a better situation -- especially since we need a root resigning in approx the next 60 days as well.
upgraded the minimum stack version to 2.7.5 to avoid
dcoutts commentedon Jun 2, 2022
@ketzacoatl you're not wrong! 😃
tp-woven commentedon Jun 3, 2022
Thanks for fixing this, I can confirm that it works for me now without the workaround. :)
avdv commentedon Jun 10, 2022
Since this issue is still open and it was not obvious to me, please note that the checks indeed have been disabled in newer Stack versions, by commit commercialhaskell/stack@4d54687 which was released in Stack 2.1.3.
If you come here because you're using Stack, you might see this error because you're using an old version.
jkoppel commentedon Jun 10, 2022
The machine I was working on has been taken down, but I both had a fresh install of Stack and had run stack upgrade.
(Although I don't remember if
stack upgradejust outright failed because of this issue.)upgraded the minimum stack version to 2.7.5 to avoid
Kleidukos commentedon Oct 28, 2022
@gbaz shall we close this issue?
istathar commentedon Oct 28, 2022
Have the key holders been updated and the process to ensure keys are rotated been transferred to Haskell Foundation yet?
Kleidukos commentedon Oct 30, 2022
@istathar The Haskell Foundation executive is now responsible for coordinating the process, although the process itself has not be reformed yet.