On Sat, Jul 05, 2014 at 10:13:08AM -0400, Carlos O'Donell wrote:
The question of corrupt DSOs and hardening against them is entirely
up to our good judgement. IIRC this patch was to harden against
these DSOs. If we think there is a real bug we should file an issue
and investigate.
Hardening against corrupt DSOs is a valid case, but the problem here
is that we don't *really* know why the fix works. In fact, I suspect
it only papers over the crashes and for any use that enables debug,
the crash will resurface. If it doesn't, then again I suspect that
the problem may be something else altogether whose symptom got fixed
somehow with this patch. Either way, there's nothing to indicate why
the fix works and why it is correct.
Your call.
I'll revert this patch. I don't have nvidia hardware to see if the
crash still persists but if it does (that is someone reports it), then
we can take a look at it afresh and figure out the best possible fix,
this time with better documentation of what we're fixing than
"Horrible workaround for horribly broken software".
Siddhesh