STR: 1. With a mail sent in the last few days selected, press "A" to archive it 2. Note that it's in the 2009-02 archive folder 3. Ctrl+Z to undo archiving, go back to the folder where it was 4. Note that while it still sorts correctly by the actual date, the date in the threadpane displays whatever the offset to your timezone causes for a date of 01/01/1970. 5. Press A to archive again, note that it's now in the (for Pacific Time) 1969-12 folder.

ouch, imap or pop3?

POP3, Local Folders (both for where I tested, and where Michael originally reported it as half of bug 479101).

taking - this applies to undo of pop3 moves/deletes in general, so it's pretty ugly. If I found a quick fix, I think it would be worth putting in b2. I suspect the general problem is a regression, but I don't know for sure.

(I'd actually take r/sr from whoever gets to this first...) The issue was that we were getting a cleared header object from a cache, but m_initedValues was set. I'll track down where that's happening and see if I can prevent that, but this fix is the right thing to do, as described in the comment. This should be a safe enough fix for b2. It passes the test suite...

I think the hdr is actually ok in the cache - when we remove a header from a db, we cut all the columns, and set m_initedValue to 0 - but if someone is holding onto that header and tries to get any values out of it, they'll get empty values because we've cut all the mork columns, but it will set m_initedValues - that's arguably the correct behavior on the part of the hdr because we treat the absence of a column/cell as an empty value. But the caller shouldn't be doing stuff with a header we've removed from the db. I'll figure out who that is.

this regression was caused by turning on correct threading - that code was trying to remove the references of a message from its hash tables after all the columns were cut from the hdr, which could not have worked very well (my bad on that code review :-( ) Sometimes we would have cached the references, and everything would be OK, but in general, not.

I'm going to put this in b2 in light of the new info, but clarkbw can overrule, of course ;-)

marking dependency on the bug that caused the regression...

Comment on attachment 363185 [details] [diff] [review] now with correct threading Stealing sr since Standard8's busy and this is on the critical path for b2.

Comment on attachment 363185 [details] [diff] [review] now with correct threading Looks good; sr=dmose

fix checked in - http://hg.mozilla.org/comm-central/rev/3e7aab1093d3

Landed on release branch: changeset: 2042:6196fbd1150a branch: COMM_20090224_RELBRANCH user: David Bienvenu <bienvenu@nventure.com> date: Fri Feb 20 12:41:02 2009 -0800 summary: fix undo of move/delete/archive of pop3 msgs, date was set to 0, r=neil, sr=dmose, a=clarkbw (as was prev push) 479163