219188 - Getting mails via POP(3) broken

Status: VERIFIED FIXED

(MailNews Core :: Networking: POP, defect)

Description

[Frank Wein [:mcsmurf]]

In recent nightlies and/or CVS builds getting mails via POP broke. When you
click on Get Msgs, Mozilla gets the mails, saves them to the mbox files and then
hangs. This is caused by the checkin of Bug 107883, backing out this changes
makes MailNews work again. I've attached a valgrind log, because it hangs at
mallopt():
#0  0x404d32c5 in mallopt () from /lib/libc.so.6
#1  0x404d22ca in free () from /lib/libc.so.6
#2  0x400c2252 in PR_Free (ptr=0x8e60640) at
/mozilla2/mozilla/nsprpub/pr/src/malloc/prmem.c:456
#3  0x41eb79fb in FreeUidlInfo (he=0x90fbc60, flag=1) at
/mozilla2/mozilla/mailnews/local/src/nsPop3Protocol.cpp:191
#4  0x400a5ae8 in PL_HashTableRawRemove (ht=0x8f735e0, hep=0xbffff108,
he=0x90fbc60) at /mozilla2/mozilla/nsprpub/lib/ds/plhash.c:330
#5  0x400a5e4c in PL_HashTableEnumerateEntries (ht=0x8f735e0, f=0x41eb7808
<net_pop3_remove_messages_marked_delete>, arg=0x8f1d9f8) at
/mozilla2/mozilla/nsprpub/lib/ds/plhash.c:449
[...]

Comment 1

[Frank Wein [:mcsmurf]]

Attachment: valgrind log

Comment 2

[Frank Wein [:mcsmurf]]

Attachment: valgrind log2

This is the right one, the first had log lines, that didn't belong to this bug

Comment 3

[David :Bienvenu]

could you attach your popstate.dat, or e-mail it to me? And your prefs.js, too?
I've been running with this patch for a month now w/o a problem, so it's
probably something specific to your configuration.

Comment 4

[Frank Wein [:mcsmurf]]

my popstate.dat is empty, do you mean my popstate.dat when it just hangs?

Comment 5

[Frank Wein [:mcsmurf]]

Attachment: my prefs.js

Comment 6

[Mats Palmgren (inactive)]

I'm also having frequent hangs using 2003-09-13-22 on Linux.
This did not occur with build 2003-09-13-05.

Comment 7

[Frank Wein [:mcsmurf]]

btw: i know that some path are wrong (for the news-server), but that hasn't
anything to do with this bug, has it? I've also from two people in newsgroups
which have the same problem with a recent nightly, so i don't seem to be the ony
one who has this problem.

Comment 8

[Hartmut Figge (don't use this one, it's defunct!)]

I was bitten by this bug with 'checkout start: Sa Sep 13 20:15:28 CEST 2003'.
Now i am back to 20030911, which works properly.

Comment 9

[David :Bienvenu]

not in 1.5, so it can't block 1.5 - clearing.

Your popstate.dat is empty? Do you have leave on server set? It looks like it's
not, from your prefs.js...

Comment 10

[David :Bienvenu]

Attachment: proposed fix

sorry about that - this should fix it.

Comment 11

[David :Bienvenu]

fix checked in, r/sr=mscott

Comment 12

[David :Bienvenu]

duh, that can't be it. This is a double free, somehow - will run purify

Comment 13

[David :Bienvenu]

Attachment: proposed fix

the deleting of the uidl info happens in the hash table callback now, so I
removed this code that does a delete.

Comment 14

[David :Bienvenu]

fix checked in, r/sr=sspitzer

Comment 15

[David :Bienvenu]

*** Bug 219213 has been marked as a duplicate of this bug. ***

Comment 16

[Frank Wein [:mcsmurf]]

Also this bug is fixed: No, i have not leave on server turned on. I'll try to
verify this bugfix today.

Comment 17

[Frank Wein [:mcsmurf]]

ok this bugfix fixed this bug, thanks :)!
V.

Comment 18

[Mats Palmgren (inactive)]

Yep, works fine for me too.

David, many thanks for your quick fixes for this bug and bug 219185, I mean
which company can deliver fixes within hours of being reported on a Sunday?!

Comment 19

[Jerry Baker]

See bug 219657 and bug 219449. This is AFTER this fix is checked in. Something
is still awry. Reopen???

Comment 20

[David :Bienvenu]

bug 219449 was filed on 1.5b - the change that caused this bug was only in 1.6,
as well as the fix for this bug. So they're completely different bugs, so please
don't re-open this one. For one thing, those bugs are about get new mail
stopping working after an hour - this bug was for get new mail not working at all.

Comment 21

[David :Bienvenu]

actually, it might just be that the trunk hasn't updated the version string -
but anyway, these are different bugs...