358559 - R_386_PC32 relocations in libxul.so when building on FC6

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

When building on Fedora Core 6 (where our system wrappers stuff works, and
where the big gcc visibility bug is fixed), SELinux prevents me from starting
due to the bad type of relocations in libtracemalloc.so and libxul.so.  This
bug covers the problem in libxul.so.

I don't understand this problem yet, so if you do, feel free to take it, but I'm planning to debug it later today.

The problem is a single relocation (as shown by objdump -R libxul.so):
0000de67 R_386_PC32        __cxa_pure_virtual

There are plenty of other relocations for __cxa_pure_virtual of the healthier R_386_32 sort, but there's this lonely one relocation that prevents my default build from running with all the SELinux protection enabled.

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

For what it's worth, objdump -sdR libxul.so shows the relocation in question as being in the midst of code compiled from toolkit/xre/nsAppRunner.cpp .  It's in:

0000de60 <__pure_virtual>:
    de60:       55                      push   %ebp
    de61:       89 e5                   mov    %esp,%ebp
    de63:       83 ec 08                sub    $0x8,%esp
    de66:       e8 fc ff ff ff          call   de67 <__pure_virtual+0x7>
                        de67: R_386_PC32        __cxa_pure_virtual
    de6b:       c9                      leave  
    de6c:       c3                      ret    
    de6d:       90                      nop    
    de6e:       66 90                   xchg   %ax,%ax

which immediately follows the definition of _ZN12nsXULAppInfo9GetVendorER19nsACString_internal (which is the last of a bunch of the nsXULAppInfo methods that are in the binary in reverse order from their order in the source).  Following it are lots of things that would normally be inlined (this is a -g -O2 -fno-inline build).

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Oh, this is the MOZ_ENABLE_OLD_ABI_COMPAT_WRAPPERS code.

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Ah, we have a system wrapper for pk11func.h but not for pk11pub.h .

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Er, wrong bug.

Comment 6

[Benjamin Smedberg]

Comment on attachment 243936 [details] [diff] [review]
patch

What were these symbols for? Can they be removed yet?

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

They're from bug 154206 -- and I'd sort of expect they're still needed.

Comment 8

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

And I realize I should land the same patch in xpfe/bootstrap, since this code still lives there too...

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Checked in to trunk.

Comment 10

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch landed

This is just the previous patch doubled, and what I landed just now.

Comment 11

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 243987 [details] [diff] [review]
patch landed

This is a low risk patch that will, along with the patches to bug 358558 and bug 358565:
 * allow Linux distros to take advantage of performance  optimizations (prelinking) and security improvements (SELinux)
 * make it easier for developers to build on new distributions like Fedora Core 6

Comment 12

[Benjamin Smedberg]

Comment on attachment 243987 [details] [diff] [review]
patch landed

FWIW, we could probably have used HAVE_VISIBILITY_ATTRIBUTE instead of doing -DWRAP_SYSTEM_INCLUDES.

Comment 13

[Daniel Veditz [:dveditz]]

Comment on attachment 243987 [details] [diff] [review]
patch landed

approved for 1.8/1.8.0 branches, a=dveditz for drivers

Comment 14

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Fix checked in to MOZILLA_1_8_0_BRANCH and MOZILLA_1_8_BRANCH.