510657 - OS/2 build break in nanojit/avmplus after checkin for bug504462

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Walter Meinl]

User-Agent:       Mozilla/5.0 (OS/2; U; Warp 4.5; en-US; rv:1.9.1.2) Gecko/20090802 Firefox/3.5.2
Build Identifier: 

E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp: In function 'void VMPI_setPageProtection(void*, size_t, bool, bool)':
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp:118: error: '_SC_PAGESIZE' was not declared in this scope
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp:118: error: 'sysconf' was not declared in this scope
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp:125: error: 'PROT_READ' was not declared in this scope
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp:127: error: 'PROT_EXEC' was not declared in this scope
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp:130: error: 'PROT_WRITE' was not declared in this scope
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp:132: error: 'mprotect' was not declared in this scope
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp: In member function 'void nanojit::CodeAlloc::freeCodeChunk(void*, size_t)':
E:/usr/src/hg/mozilla-central/js/src/nanojit/avmplus.cpp:181: error: no matching function for call to 'nanojit::CodeAlloc::free(void*&)'
E:/usr/src/hg/mozilla-central/js/src/nanojit/CodeAlloc.h:159: note: candidates are: void nanojit::CodeAlloc::free(nanojit::NIns*, nanojit::NIns*)

A workaround for the moment is to build with --disable-jit

Reproducible: Always

Comment 1

[Rich Walsh]

Attachment: trivial fixes for breakage [checked in comment 9]

Comment 2

[Peter Weilbacher]

Interesting. If including mman.h solves our problem does that mean we do have a working mmap in kLibc? That would also significantly improve our chances to fix bug 510705 and bug 417450.

Comment 3

[Dave Yeo]

No working mmap in klibc, just the headers.
I've tried linking with Yuri's mmap.lib but it still crashes. I'll have to review the build log to make sure I didn't screw up

Comment 4

[Graydon Hoare :graydon]

mmap is not required here, so much as "a way to get some memory we can switch the executable protection bits on". valloc() is the fallback, you might want to try that?

Afraid I'm not much of an OS/2 expert, don't know what it exposes.

Comment 5

[Graydon Hoare :graydon]

Comment on attachment 394754 [details] [diff] [review]
trivial fixes for breakage [checked in comment 9]

This is fine if it works, but I'd be surprised if it actually *does* if OS/2 lacks mmap. This will make it build, but does it run?

Comment 6

[Rich Walsh]

(In reply to comment #5)
> This is fine if it works, but I'd be surprised if it actually *does* if
> OS/2 lacks mmap. This will make it build, but does it run?

Sure, no problems.  We had to add sys/mman.h to get access to mprotect() which on OS/2 is just a thin wrapper around an OS API.  valloc() also works as it should, returning page-aligned memory.

However, the region of memory valloc() uses could be problematic in some circumstances.  It might be safer if we substituted platform-specific code for both functions (much like the Win code).

Comment 7

[Peter Weilbacher]

Should we still get the trivial fix in as an intermediate solution?

Comment 8

[Andreas Gal :gal]

Absolutely. Please assign to me if you want me to land it, otherwise any volunteer will do.

Comment 9

[Peter Weilbacher]

OK: http://hg.mozilla.org/mozilla-central/rev/ccc83c30337a

Comment 10

[Rich Walsh]

Attachment: OS/2-specific code for avmplus

This adds platform-specific code to allocate, free, and set page protection for nanojit code chunks.  My earlier patch which fixed a build breakage remains valid.

Comment 11

[Rich Walsh]

Attachment: OS/2 code for avmplus - v2 [checked in comment 13]

Comment 12

[Peter Weilbacher]

Comment on attachment 396151 [details] [diff] [review]
OS/2 code for avmplus - v2 [checked in comment 13]

Looks good from an OS/2 POV, but I guess a JS person should review this.

Andreas, can you do that?

Comment 13

[Peter Weilbacher]

Great, thanks again! Pushed to mozilla-central:
http://hg.mozilla.org/mozilla-central/rev/45c8ca4ab959

Comment 14

[Robert Sayre]

http://hg.mozilla.org/releases/mozilla-1.9.2/rev/766a7053e76b

Comment 15

[Robert Sayre]

http://hg.mozilla.org/releases/mozilla-1.9.2/rev/64a265f676bd