265536 - PFS returns x86 plugin for AMD64 browser

Status: RESOLVED FIXED

(Core :: Networking: HTTP, defect)

Description

[Nicholas Miell]

User-Agent:       Mozilla/5.0 (X11; U; Linux x86_64; rv:1.7.3) Gecko/20041020 Firefox/0.10.1
Build Identifier: Mozilla/5.0 (X11; U; Linux x86_64; rv:1.7.3) Gecko/20041020 Firefox/0.10.1

The Plugin Finder Service ignores the architecture that the end user is using,
and  returns inappropriate plugins because of this. 

i.e. on a page which has a Flash object, the PFS finds Macromedia Flash for
Linux, despite the fact that Flash for Linux is only available for x86, not x86-64.

Attempting to intall it fails (obviously).

Reproducible: Always
Steps to Reproduce:
1. Use Firefox on Linux AMD64.
2. Find a page with a Flash object.
3. Use the PFS.

Actual Results:  
PFS suggests an x86 plugin.

Expected Results:  
PFS correctly deduces that no plugin is available.

Comment 1

[Johnny Stenback (:jst)]

Yes, known problem. The problem is that we can't distinguish between those who
run a 32-bit version of Firefox and those who run a 64-bit version, and we don't
want to not give out the 32-bit flash player to the users that can use it. Users
of 64-bit versions of Firefox (including me) simply don't get a usable flash
player, which just doesn't work (doesn't crash the browser). Once one is
available, it'd be great if we could feed the right thing, but I doubt this will
change for Firefox 1.0.

Comment 2

[Doron Rosenberg (IBM)]

*** Bug 269496 has been marked as a duplicate of this bug. ***

Comment 3

[Mike Hommey [:glandium]]

Apparently, it also does get the i386 flash plugin on alpha:

p2@fred:~/.firefox/plugins$ file libflashplayer.so
libflashplayer.so: ELF 32-bit LSB shared object, Intel 80386, version 1 (SYSV),
stripped
p2@fred:~/.firefox/plugins$ uname -a
p2-mate Linux fred 2.6.3 #6 Tue Jun 22 16:52:10 CEST 2004 alpha GNU/Linux

[note that the test on alpha is courtesy of Peter De Schrijver]

Comment 4

[David Bélanger]

Gets the Linux/x86 plugin on Linux/PowerPC also.  Reports install successful...  

There is currently no Linux/PowerPC Flash binary.


David

Comment 5

[timeless]

[e-mail]: that should be a different bug (which i believe is filed)

Comment 6

[Doron Rosenberg (IBM)]

*** Bug 284313 has been marked as a duplicate of this bug. ***

Comment 7

[Johnny Stenback (:jst)]

I changed the service to only offer the flash installer to Linux i<n>86 systems.

Comment 8

[Johnny Stenback (:jst)]

Attachment: Make the UA string tell 32 and 64 bit apps appart on x86_64.

Comment 9

[Doron Rosenberg (IBM)]

Would this something we would want for a Firefox 1.0.2?

Comment 10

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 176092 [details] [diff] [review]
Make the UA string tell 32 and 64 bit apps appart on x86_64.

r=bzbarsky

Comment 11

[Darin Fisher]

Comment on attachment 176092 [details] [diff] [review]
Make the UA string tell 32 and 64 bit apps appart on x86_64.

I think you should test sizeof(void*) instead since that's really what you
mean.  I can imagine some funky configuration where long is still 32-bits. 
sr=darin

Comment 12

[Johnny Stenback (:jst)]

Made that change and landed this. FIXED.

Comment 13

[Johnny Stenback (:jst)]

Yeah, IMO we should take this if we do a 1.0.2.

Comment 14

[Daniel Veditz [:dveditz]]

Ought to be fixed, but not blocking 1.0.3

Comment 15

[Mike Shaver (:shaver emeritus)]

Comment on attachment 176092 [details] [diff] [review]
Make the UA string tell 32 and 64 bit apps appart on x86_64.

a=shaver.  (As a Linux/x86-64 user, I hate this bug.)

That comment doesn't make sense, though: the (x86_64) doesn't distinguish from
the 64-on-64 case, it's the i686 that does so, no?

Comment 16

[Daniel Veditz [:dveditz]]

Probably wouldn't actually hold for this, please land soon.

Comment 17

[Johnny Stenback (:jst)]

Attachment: Aviary branch version.

Comment 18

[Johnny Stenback (:jst)]

Moving bug to Core/http since this changes the UA string and should also be
fixed for 1.7...

Comment 19

[Jay Patel [:jay]]

Comment on attachment 176092 [details] [diff] [review]
Make the UA string tell 32 and 64 bit apps appart on x86_64.

Let's get this in for 1.7.9  while we're at it. a=jay

Comment 20

[Johnny Stenback (:jst)]

Landed on the 1.7 branch.