On Tue, 2018-02-20 at 09:18 -0600, Steve Wise wrote:
> >
> > On Mon, 2018-02-19 at 22:53 -0500, Doug Ledford wrote:
> > > On Mon, 2018-02-19 at 16:11 -0700, Jason Gunthorpe wrote:
> > > > On Mon, Feb 19, 2018 at 05:09:15PM -0600, Steve Wise wrote:
> > > >
> > > > > > And please check patchworks, there is something wrong with how you
> >
> > send
> > > > > > patches they get wronly ordered and have wonky dates.. Makes them
> >
> > hard
> > > > > > to apply..
> > > > >
> > > > > I'm sending them with sendmail. I'll figure it out...
> > > >
> > > > So do I, I just have git-send-email call it for me :)
> > > >
> > > > [sendemail]
> > > > smtpserver = /usr/sbin/sendmail
> > > > confirm = always
> > > > from = Jason Gunthorpe <jgg@xxxxxxxx>
> > > > envelopeSender = Jason Gunthorpe <jgg@xxxxxxxx>
> > > > suppresscc = self
> > > >
> > > > Jason
> > >
> > > I think it's the size of the patch. I've seen this before. Big patches
> > > can end up landing in patchworks out of order, and I think it's because
> > > the mailman queues on vger simply send the files, so smaller patches,
> > > even though queued later, can end up ahead of bigger patches because
> > > they deliver to recipients faster.
> > >
> >
> > Ok, after looking, none of these patches are large enough to cause the
> > effect I talked about. It's normally things like 10,000 line binary
> > blob patches that get obviously out of sequence. Everything here is
> > small enough I would expect it to deliver in order, so I don't know
> > what's going on here except maybe the specific smtp server Steve is
> > using might be reordering things?
> >
> > There is another option, but it only applies to direct and list
> > recipients such as myself and Jason. I filter my incoming email to
> > remove duplicates. Sometimes the list email arrives first and sometimes
> > the direct email arrives first. Just depends. Either way, they are all
> > filtered to the same folder and de-duped (which means if you *really*
> > want to get my attention, you email me without including a mailing
> > list).
> >
> > But, that's not it either. Looking at this thread, the cover letter is
> > from Steve at 2:15pm. However, patch 2/9 is from Steve at 11:16am. See
> > for yourself:
> >
> > Cover Letter headers:
> > From: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
> > Date: Mon, 19 Feb 2018 11:15:02 -0800 (02/19/2018 02:15:02 PM)
> > Subject: [PATCH v2 rdma-next 0/9] cm_id, cq, mr, and pd resource
> > tracking
> >
> > Patch 2/9 headers:
> > From: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
> > Date: Mon, 19 Feb 2018 08:16:33 -0800 (02/19/2018 11:16:33 AM)
> > Subject: [PATCH v2 rdma-next 2/9] RDMA/CM: move rdma_id_private to
> > cma_priv.h
> >
> >
> Oh I think I know what's happening. I created these patches on a machine that is in PST timezone, and the clock on the system might not even be setup correctly. I use git format-patch for that. I then pull the patches to Texas (CST timezone) and submit them to my email server via sendmail. So the "Date: " in the patch is from the PST patch creation time, and the "Received: " times logged by ogc are CST.
>
> Why is this confusing patchworks though? If it uses the patch dates only or the received dates only, it should be in-order, I think.
Even that doesn't explain the above. There are two vastly different
timestamps on the emails. Patchworks is evidently using the Date:
header, and on the two different files coming from the computer in the
PST timezone, the times are different by 3 hours.
--
Doug Ledford <dledford@xxxxxxxxxx>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDDAttachment:
signature.asc
Description: This is a digitally signed message part
