RE: [GHC] #16348: GHC HEAD regression: tyConAppArgs

Matthew, Ben
I've just received 40-odd messages like this one. It looks as if Marge is now sending commit messages at Trac ticket messages, which is great. Will that happen after the move to GitLab.
Also, is this sudden wave because a whole lot of commits have now landed in master? Or is it somehow an old backlog stuck in a mail queue?
There may be some housekeeping to do, to close tickets, check regression tests and add pointers to the appropriate tests. Is anyone up for doing that?
Thanks
Simon
| -----Original Message-----
| From: ghc-tickets

Simon Peyton Jones via ghc-devs
Matthew, Ben
I've just received 40-odd messages like this one. It looks as if Marge is now sending commit messages at Trac ticket messages, which is great. Will that happen after the move to GitLab.
Also, is this sudden wave because a whole lot of commits have now landed in master? Or is it somehow an old backlog stuck in a mail queue?
This happened because the GitLab -> git.haskell.org mirroring process was stuck. Yesterday I un-stuck it which then triggered the push of approximately 40 commits, triggering the old Trac commit notifier which produced the messages you received. Regarding commit notifications after we migrate: * Tickets will have notes added when they are mentioned by a commit message. As we discussed earlier, messages won't include the commit message text but rather only a reference to the referring commit SHA. For instance, this looks like [1]. If we find the indirection between the ticket and the mentioning commit message to be problematic we can certainly revisit this. * Commit notifications will be sent to ghc-commits@haskell.org; the format will change a bit but the overall content won't change. * Unfortunately a mention of a ticket from a commit does not produce a notification email. This is in my opinion a rather serious issue that we will need to work around since it makes closing tickets after merge far more painful than necessary.
There may be some housekeeping to do, to close tickets, check regression tests and add pointers to the appropriate tests. Is anyone up for doing that?
Yes, I have been accumulating a sizeable queue of tickets to sort through. I've started working through this but certainly won't finish tonight. I have a few other obligations tomorrow but I'll try to pick it up again later in the day. Cheers, - Ben [1] https://gitlab.staging.haskell.org/ghc/ghc/issues/16260#note_173847

Ben wrote:
Unfortunately a mention of a ticket from a commit does not produce a notification email. This is in my opinion a rather serious issue that we will need to work around since it makes closing tickets after merge far more painful than necessary.
Is there something in the GHC workflow that precludes the normal Gitlab
behavior of automatically closing issues after a merge? There just needs to
be a commit message or MR description with the words "Closes #xxx" [1].
Or perhaps you were talking specifically about Trac?
I'm just spectating this from a distance so might have misunderstood the
problem.
[1]: Full details at
https://docs.gitlab.com/ee/user/project/issues/closing_issues.html#via-merge...
On Wed, 6 Mar 2019, 5.57 Ben Gamari,
Simon Peyton Jones via ghc-devs
writes: Matthew, Ben
I've just received 40-odd messages like this one. It looks as if Marge is now sending commit messages at Trac ticket messages, which is great. Will that happen after the move to GitLab.
Also, is this sudden wave because a whole lot of commits have now landed in master? Or is it somehow an old backlog stuck in a mail queue?
This happened because the GitLab -> git.haskell.org mirroring process was stuck. Yesterday I un-stuck it which then triggered the push of approximately 40 commits, triggering the old Trac commit notifier which produced the messages you received.
Regarding commit notifications after we migrate:
* Tickets will have notes added when they are mentioned by a commit message. As we discussed earlier, messages won't include the commit message text but rather only a reference to the referring commit SHA. For instance, this looks like [1].
If we find the indirection between the ticket and the mentioning commit message to be problematic we can certainly revisit this.
* Commit notifications will be sent to ghc-commits@haskell.org; the format will change a bit but the overall content won't change.
* Unfortunately a mention of a ticket from a commit does not produce a notification email. This is in my opinion a rather serious issue that we will need to work around since it makes closing tickets after merge far more painful than necessary.
There may be some housekeeping to do, to close tickets, check regression tests and add pointers to the appropriate tests. Is anyone up for doing that?
Yes, I have been accumulating a sizeable queue of tickets to sort through. I've started working through this but certainly won't finish tonight. I have a few other obligations tomorrow but I'll try to pick it up again later in the day.
Cheers,
- Ben
[1] https://gitlab.staging.haskell.org/ghc/ghc/issues/16260#note_173847 _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

On March 6, 2019 2:06:25 AM EST, Bryan Richter wrote:
Ben wrote:
Unfortunately a mention of a ticket from a commit does not produce a notification email. This is in my opinion a rather serious issue that we will need to work around since it makes closing tickets after merge far more painful than necessary.
Is there something in the GHC workflow that precludes the normal Gitlab behavior of automatically closing issues after a merge? There just needs to be a commit message or MR description with the words "Closes #xxx" [1].
Or perhaps you were talking specifically about Trac?
I'm just spectating this from a distance so might have misunderstood the problem.
[1]: Full details at https://docs.gitlab.com/ee/user/project/issues/closing_issues.html#via-merge...
The problem is that you very often want to review the ticket before closing it. There might be several reasons for this: - You might believe that your commit closes the ticket but perhaps there is a facet of the problem that you are forgetting - You might want to write a final summary of the solution - Perhaps there is an additional refactoring that you thought of while writing your patch that you would like to open a new ticket to track. - Maybe you just want to mark the patch to be back ported to the stable branch. The general theme here is that looking over the ticket one last time before closing is a helpful practice that allows us to catch problems early and keep our issue tracker tidy. It would be quite unfortunate to lose this ability. Cheers, - Ben
On Wed, 6 Mar 2019, 5.57 Ben Gamari,
wrote: Simon Peyton Jones via ghc-devs
writes: Matthew, Ben
I've just received 40-odd messages like this one. It looks as if Marge is now sending commit messages at Trac ticket messages, which is great. Will that happen after the move to GitLab.
Also, is this sudden wave because a whole lot of commits have now landed in master? Or is it somehow an old backlog stuck in a mail queue?
This happened because the GitLab -> git.haskell.org mirroring process was stuck. Yesterday I un-stuck it which then triggered the push of approximately 40 commits, triggering the old Trac commit notifier which produced the messages you received.
Regarding commit notifications after we migrate:
* Tickets will have notes added when they are mentioned by a commit message. As we discussed earlier, messages won't include the commit message text but rather only a reference to the referring commit SHA. For instance, this looks like [1].
If we find the indirection between the ticket and the mentioning commit message to be problematic we can certainly revisit this.
* Commit notifications will be sent to ghc-commits@haskell.org; the format will change a bit but the overall content won't change.
* Unfortunately a mention of a ticket from a commit does not produce a notification email. This is in my opinion a rather serious issue that we will need to work around since it makes closing tickets after merge far more painful than necessary.
There may be some housekeeping to do, to close tickets, check regression tests and add pointers to the appropriate tests. Is anyone up for doing that?
Yes, I have been accumulating a sizeable queue of tickets to sort through. I've started working through this but certainly won't finish tonight. I have a few other obligations tomorrow but I'll try to pick it up again later in the day.
Cheers,
- Ben
[1] https://gitlab.staging.haskell.org/ghc/ghc/issues/16260#note_173847 _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
-- Sent from my Android device with K-9 Mail. Please excuse my brevity.

| * Tickets will have notes added when they are mentioned by a commit
| message. As we discussed earlier, messages won't include the commit
| message text but rather only a reference to the referring commit SHA.
| For instance, this looks like [1].
I thought we'd decided, for now anyway, to /include/ the commit message text. (Matthew dissented.) Is that hard to do?
Simon
| -----Original Message-----
| From: Ben Gamari

On March 6, 2019 4:23:37 AM EST, Simon Peyton Jones
| * Tickets will have notes added when they are mentioned by a commit | message. As we discussed earlier, messages won't include the commit | message text but rather only a reference to the referring commit SHA. | For instance, this looks like [1].
I thought we'd decided, for now anyway, to /include/ the commit message text. (Matthew dissented.) Is that hard to do?
Simon
Oh dear, I will need to look back over my notes. I suspect my memory has failed me here. Regardless, it is not particularly hard to include commit messages. Cheers, - Ben -- Sent from my Android device with K-9 Mail. Please excuse my brevity.
participants (3)
-
Ben Gamari
-
Bryan Richter
-
Simon Peyton Jones