Open Bug 638936 Opened 11 years ago Updated 2 months ago

Only use two last numbers in NVIDIA driver version numbers


(Core :: Graphics, defect)

Windows 7
Not set



Tracking Status
blocking2.0 --- .x+


(Reporter: bjacob, Unassigned)




(1 file, 1 obsolete file)

A user with an old NVIDIA Quadro driver on Windows XP is reporting he has driver version, so he's getting unwillingly whitelisted because on XP we require (while on vista/7 we require

Let's only use the two last numbers here (12.5721).
Tryserver build:

Not suggesting that we absolutely need this in Firefox This patch is not 100% risk-free. I'm testing it as much as I can by spoofing, but my advice is to take that in the first bugfix update.
Attachment #517010 - Flags: review?(joe)
I have now tested this patch a lot using spoofing, and it seems to be running well.

Also, I have extracted some stats from crash reports, and they hint that we should really use this patch:

$ grep AdapterVendorID 20110306-pub-crashdata.csv | wc -l

That tells us that 94.4K reports received yesterday were Firefox 4 (since Firefox 3 reports don't have AdapterVendorID)

$ grep AdapterVendorID\:\ 10de 20110306-pub-crashdata.csv | wc -l

That tells us that out of these 94.4K reports, at least 29.6K are using Nvidia cards.

$ egrep AdapterDriverVersion\:\ [0-9]\\.[0-9]\\.[0-9]\\.[0-9] 20110306-pub-cras
hdata.csv  | grep AdapterVendorID\:\ 10de | wc -l

That tells us that out of these 29.6K Nvidia reports, 5.5K have these weird version numbers. That's 19% of all reports using Nvidia cards.

$ egrep AdapterDriverVersion\:\ [7-9]\\.[0-9]\\.[0-9]\\.[0-9] 20110306-pub-cras
hdata.csv  | grep AdapterVendorID\:\ 10de | wc -l

That tells us that out of these 5.5K reports, 1.7K are on machines getting incorrectly whitelisted (assuming they're all on WindowsXP), like the version '' initially reported, because >=

1735/29621 = 6% of Nvidia users
1735/94621 = 1.8% of all Firefox 4 reports

So I think that's quite serious and we should land the patch.
blocking2.0: --- → ?
There actually was a bug there that would have made the current entry in the downloaded blocklist, using 8.17.x.x numbers, not take effect. Scary! patch updated.
Attachment #517010 - Attachment is obsolete: true
Attachment #517010 - Flags: review?(joe)
Attachment #517446 - Flags: review?(bjacob)
Attachment #517446 - Flags: review?(bjacob) → review?(joe)
After discussion with Joe: since:
 * this is important enough that we want it in Firefox 4.0 if possible, but
 * this is probably not important enough to delay Firefox 4 release by itself,
this looks like a good 'RC ride-along' candidate.
Comment on attachment 517446 [details] [diff] [review]
Don't get confused by NVIDIA driver versions, updated

As follow-ups for this, we should a) insert a test that makes sure stuff is blocked on XP, and that the first two numbers also don't impact the blacklisting; b) contact NVIDIA to figure out what their long-term driver version style is going to be, so we can code for that.
Attachment #517446 - Flags: review?(joe) → review+
Er, actually, maybe we should remove that const_cast<>, and use a temporary instead. Not sure whether we can rely on the ability to modify const globals.
You can see the tests I wrote in bug 625160, and use them as a template.
blocking2.0: ? → .x+
Assignee: nobody → bjacob

The bug assignee didn't login in Bugzilla in the last 7 months.
:bhood, could you have a look please?
For more information, please visit auto_nag documentation.

Assignee: jacob.benoit.1 → nobody
Flags: needinfo?(bhood)
Flags: needinfo?(bhood)
You need to log in before you can comment on or make changes to this bug.