Dominique Martinet
changed
bug 8787
| What |
Removed |
Added |
| Attachment #11173 is obsolete |
|
1
|
| Attachment #11193 Flags |
|
review_for_checkin?
|
Comment # 16
on bug 8787
from Dominique Martinet
Created attachment 11193 [details]
[PATCH] build fids on first pass, v4
Okay, here's a new one with some fixes.
(T/RVERSION shouldn't set tag as they use it for the actual version enum; and
fid is modified on create only on reply as a fix for what I was talking about -
this ends up allocating one se_alloc string that will be used as is on success,
but which ends up lost on error. I supposed you'd prefer this to a gmalloc'd
string that might end up never freed if there is no coherent reply at all -
correct me if I'm wrong)
I also just found out that in the sample I gave you in the previous issue,
there also was (at least) one 9P message cut in two TCP packets... Which is
perfectly legal, but ends up completely misinterpreted. Most end up recognized
as junk, but I noticed this one because the "version" field was wrong for the
whole conversation because of this one :/
As I said this does not utterly bother me as it is pretty rare (and I don't
usually look at tcp messages anyway, but that's a tad more complicated although
it provides me with perfectly-cut packets :D), but I'd think this is worth
investigating if there is anything commonly done in other dissectors.
Will gladly fix any more style issues if there's more wrong :)
You are receiving this mail because:
- You are watching all bug changes.