511227 - Firefox 3.0.13 fails to compile on FreeBSD/powerpc

Status: VERIFIED FIXED

(NSS :: Build, defect, P1)

Description

[chmeeedalf]

User-Agent:       Mozilla/5.0 (X11; U; FreeBSD powerpc; en-us) AppleWebKit/532+ (KHTML, like Gecko) Safari/419.3 Midori/0.1.7
Build Identifier: Mozilla/5.0 (X11; U; FreeBSD powerpc; en-US; rv:1.9.0.13) Gecko/2009081815 Firefox/3.0.13

With the change in bug #431958 the nss library in firefox 3.0.13 does not compile.  This was tracked down to the mozilla/security/coreconf/FreeBSD.mk file, and a patch for this is attached.

Reproducible: Always

Steps to Reproduce:
1. Build the www/freebsd3 port on FreeBSD/powerpc
2. Watch it error in sha_fast.c with assembler error of unknown opcode 'bswap'
3.

Comment 1

[chmeeedalf]

Attachment: Patch to set CPUTYPE appropriately

Comment 2

[Nick Thomas [:nthomas] (UTC+12)]

Should this go live in NSS:Build or Core:Build Config ?

Comment 3

[chmeeedalf]

From a cursory look at the Makefile and FreeBSD.mk files from both nss-3.12.3 and firefox 3.0.13, the bug affects the separate NSS library as well as Firefox, so it's probably best to apply the patch to that library, and include it in the next import to firefox and other projects that import it.

Comment 4

[Nelson Bolyard (seldom reads bugmail)]

I will commit this patch on the CVS upstream repository.

Comment 5

[Wan-Teh Chang]

chmeeedalf: thanks for the patch.  Two questions:

1. Would "uname -m" work?  If so, we can just use
the $(OS_TEST) variable (which equals "uname -m").
$(OS_TEST) is what most other platforms use, such as Linux.mk:
http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/security/coreconf/Linux.mk&rev=1.41&mark=55,57,62,6466,69#55

2. Does "uname -p" actually return "amd64"?  Or did
you test "amd64" simply because I mentioned it in
bug 431958 comment 55?

Comment 6

[Wan-Teh Chang]

One more question: does "uname -p" (or "uname -m") ever
return "i686" on FreeBSD?

Re: question 1 in my previous comment, I read the uname(1)
man page on FreeBSD, and I think "uname -p" is more
appropriate for $(CPU_ARCH) whereas  "uname -m" is more
portable.  (The -p option is not listed in
http://www.opengroup.org/onlinepubs/000095399/utilities/uname.html )

Comment 7

[Wan-Teh Chang]

Did some research.  The GNU config.guess script uses
"uname -m" for FreeBSD and maps

  pc98 -> i386
  amd64 -> x86_64
  everything else -> self

http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob;f=config.guess;hb=HEAD

So we should be able to just test $(OS_TEST) (which is
"uname -m").  Not sure if we still need to handle pc98.

Comment 8

[chmeeedalf]

"uname -m" should be sufficient, yes.  I only have i386 and powerpc FreeBSD systems, and "uname -m" and "uname -p" return the same values for both on these systems.

Comment 9

[Wan-Teh Chang]

Attachment: Patch to set CPU_ARCH appropriately, v2

chmeeedalf, could you please test this patch?  Thanks.

I found the list of supported processors at
http://www.freebsd.org/releases/7.1R/hardware.html#PROC

Since "pc98" is listed in the FreeBSD 7.1 release notes,
I map treat it the same as "i386" as config.guess does.

I checked in this patch on the NSS trunk (3.12.5).

Checking in FreeBSD.mk;
/cvsroot/mozilla/security/coreconf/FreeBSD.mk,v  <--  FreeBSD.mk
new revision: 1.12; previous revision: 1.11
done

Comment 10

[Wan-Teh Chang]

chmeeedalf, do you have some NSPR patches for FreeBSD/powerpc?

NSPR's _freebsd.cfg can't be compiled as is.  Compilation will
fail here:

http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/nsprpub/pr/include/md/_freebsd.cfg&rev=3.11&mark=286,288#277

Comment 11

[chmeeedalf]

Your patch v2 worked.

There is a patch in ports that fixes NSPR for powerpc.  It can be found at

http://www.freebsd.org/cgi/cvsweb.cgi/ports/devel/nspr/files/patch-.._pr_include_md__freebsd.cfg?rev=1.3;content-type=text%2Fplain

There are a few other patches for the port as well, if you want to cherry pick them as desired.

Comment 12

[Wan-Teh Chang]

Thank you for testing my patch v2.

I filed bug 512076 to remind myself to look at the
NSPR patches in FreeBSD port.  Thanks for the link.