didier gautheron
changed
bug 9112
What |
Removed |
Added |
Status |
UNCONFIRMED
|
CONFIRMED
|
CC |
|
[email protected]
|
Ever confirmed |
|
1
|
Comment # 8
on bug 9112
from didier gautheron
Hi,
(In reply to comment #6)
> Thanks for formatting!
>
> I think that this bug came to life after the gulong -> guint32 switch done
> in:
> r1=46402&r2=46748">http://anonsvn.wireshark.org/viewvc/trunk/epan/follow.
> c?r1=46402&r2=46748
>
> About http://anonsvn.wireshark.org/viewvc?view=revision&revision=31334,
> I see here the meaning of changing the curly bracket's location, but yes
> dont understand the reason of the other line's glongcasting/subtraction
> instead of previous single '>' operator..
> Maybe Sake can answer?
I guess it's for sequence wrap around but casting with glong on 64 bits system
is wrong should be gint32 or better use macros LT_SEQ/GT_SEQ defined in
dissectors/packet-tcp.h,.
It works, sort of, without casting because most of the time wireshark is using
relative sequences and there's no wrap around (wireshark will grind to a halt
if you try to follow a > 4GB conversation :)).
>
> What I know is that, after applying my patch, I've never seen again
> incorrect "bytes missing in capture file" msgs, and all data are ok
> [I use tshark/wireshark mainly for decoding http streams, and I've done by
> myself the dechunk/gunzip routines, since I've seen they were broken in
> wireshark].
>
> If I'll have time, I'll anonymize some capture files that show the issue.
You are receiving this mail because:
- You are watching all bug changes.