Comment # 14
on bug 10798
from Hadriel Kaplan
(In reply to Alexis La Goutte from comment #13)
> I have try to open on my Mac, capture of Megaco (Bug 10898 and 10899), i
> have get the same issue...
Huh, I think that Megaco failure may show the problem the most clearly - it
fails to parse the Megaco message very early on. In fact, if I'm reading it
right, I think it shows that the tvb_previous_offset is wrong in the do-while
loop in dissect_megaco_text(), and my guess is that's because tvb_pbrk_guint8()
is returning a wrong result.
So digging into tvb_pbrk_guint8(), I see that the underlying code does
something different if the Intel CPU has the SSE4.2 extended instructions or
not. If your CPU has SSE4.2, then it will run _ws_mempbrk_sse42(), and within
that are a bunch of Intel SSE functions. Since such functions have had bugs in
GCC (and other compilers) in the past, I wonder if the bug's inside this
_ws_mempbrk_sse42() somewhere.
Perhaps I should submit a change into master repo to not call
_ws_mempbrk_sse42(), and see if the nightly build's bug goes away?
You are receiving this mail because:
- You are watching all bug changes.