181913 - NSS Build breaks on OS/2 when using GNU Make 3.79.1

Status: RESOLVED FIXED

(NSS :: Build, defect, P2)

Description

[jhp (no longer active)]

Comment 1

[jhp (no longer active)]

Attachment: patch

For some reason, the NSS build (as part of full Mozilla build) breaks when
using GNU Make 3.79.1.	This patch fixes the issue.

Comment 2

[Wan-Teh Chang]

Comment on attachment 107391 [details] [diff] [review]
patch

We need to understand what caused the build failure.
Could you provide the build log?

This patch effectively comments out the code because
MOZ_OS2_TOOLKIT is not set.  It is MOZ_OS2_TOOLS that
is used by NSS and if MOZ_OS2_TOOLS equals VACPP,
XP_OS2_VACPP is set to 1.

Comment 3

[jhp (no longer active)]

Well, that's embarassing!  Good catch.

This is the error I get:
=====================================
cd freebl; make.exe -j1 libs
make.exe[3]: Entering directory `D:/builds/trunk/mozilla/security/nss/lib/freebl
'
alp.exe -o D:/builds/trunk/mozilla/obj/nss/freebl/mpi_x86.obj +Od -c mpi/mpi_x86
.s
Licensed Materials - Property of IBM
IBM Assembly Language Processor Version 4.00.008
(C) Copyright IBM Corp., 1995-2000.  All Rights Reserved.
US Government Users Restricted Rights - Use, duplication or disclosure
restricted by GSA ADP Schedule Contract with IBM Corp.

Usage: Invalid option "-o"
make.exe[3]: *** [D:/builds/trunk/mozilla/obj/nss/freebl/mpi_x86.obj] Error 5
make.exe[3]: Leaving directory `D:/builds/trunk/mozilla/security/nss/lib/freebl'
=====================================

For some reason, it is choosing to go down the Linux path in the Makefile.  I
don't know what my patch did, but I did not get this error when I changed that
line.  Still investigating.

Comment 4

[Mike Kaply [:mkaply]]

Adding Julian

Comment 5

[jhp (no longer active)]

OBJS includes "D:/builds/trunk/mozilla/obj/nss/freebl/mpi_x86.obj".  Is it
finding 'mpi_x86.s' instead of 'mpi_x86.asm' for some reason?  How do I make it
choose one over the other?

Comment 6

[Wan-Teh Chang]

Try adding OS2 before WIN% in this line in mozilla/security/coreconf/rules.mk:

417 ifneq (,$(filter-out WIN%,$(OS_TARGET)))

Without that, foo.obj can be generated from both foo.s and foo.asm, and
foo.s is chosen because its rule is listed first.  This should have failed
the same way under older versions of GNU Make.

Comment 7

[jhp (no longer active)]

Attachment: patch 2nd try

Yes, that did it.  Don't know why this didn't fail with make 3.76.1

Comment 8

[Wan-Teh Chang]

Comment on attachment 107481 [details] [diff] [review]
patch 2nd try

Do you know if the EMX build uses the .s suffix for
assembler files?  If so, we will need to tweak this
patch.

Comment 9

[jhp (no longer active)]

The EMX build is currently broken.  There is a bug open for that, and I had just
thought of fixing this there.  But now that you mention it, I could just get
this out of the way.

Well, this is what I came up with for the if statement:
  ifneq (,$(filter-out WIN% OS2icc,$(OS_TARGET)$(CC)))
Is it too crazy?  Is there another way to test for any windows, and for the
Visual Age compiler on OS/2 all in the same if statement?

Comment 10

[Wan-Teh Chang]

I think we should just use two nested if statements.
I am afraid that there is no elegant solution to this
problem.

Comment 11

[jhp (no longer active)]

Attachment: patch for only VACPP

Comment 12

[Wan-Teh Chang]

Comment on attachment 107950 [details] [diff] [review]
patch for only VACPP

r=wtc.

Comment 13

[jhp (no longer active)]

WTC or Julien:  Can either of you please check this in?  Don't have check in
authority to NSS.  Thanks.

Comment 14

[Wan-Teh Chang]

Nelson, could you review this patch too?

Comment 15

[Nelson Bolyard (seldom reads bugmail)]

r=nelsonb 

Comment 16

[Wan-Teh Chang]

The patch has been checked into the trunk and NSS_3_7_BRANCH
of NSS.

I will mark the bug fixed when it has been checked into the
NSS_CLIENT_TAG.

Comment 17

[Wan-Teh Chang]

The fix has been checked into the NSS_CLIENT_TAG.
Marked the bug fixed.