195797 - Build NSS with GCC 3.x

Status: RESOLVED FIXED

(NSS :: Build, defect, P2)

Description

[jhp (no longer active)]

Derived from Meta bug 177789.  This bug deals specifically with patches to make
NSS build on OS/2 using GCC 3.x.

Comment 1

[jhp (no longer active)]

Attachment: patch

Comment 2

[jhp (no longer active)]

Attachment: patch v1.1

Missing endif in OS2.mk

Comment 3

[jhp (no longer active)]

Attachment: patch v1.2

More cleanup.

Comment 4

[jhp (no longer active)]

Comment on attachment 117253 [details] [diff] [review]
patch v1.2

Trying to get this approved and checked in for 1.4a

Comment 5

[Robert Relyea]

Question: is the desire to really only allow GCC to work (not CSet)? It seems
like the patch would break CSet compiles (unless there are some if's I'm missing).

bob

Comment 6

[jhp (no longer active)]

Attachment: patch v1.3

Small change to sign.cmd

Comment 7

[jhp (no longer active)]

For the moment, both GCC and Visual Age (CSet) compilers will work.  In the
future, though, we will switch to only supporting GCC.

Comment 8

[Julien Pierre]

My understanding is that the Mozilla browser needs an updated C++ compiler, and
C Set ++ / Visual Age have fallen too far behind to be supported for the build.
There have been ongoing complaints from client land to use those features, and
OS/2 was the only platform holding them back. This means that the browser will
no longer build with C Set ++ or VAC++ on OS/2.

However, NSS and NSPR currently have a C API, and there is no requirement for
them to be built with gcc. I'd like to keep the ability to build them with VAC.
I use them a lot with small programs such as the NSS command-line tools that are
C-based, even if the browser won't build with those compilers.

I'll try Javier's patches and see if my VAC build still works.

Comment 9

[Julien Pierre]

This patch didn't break my VAC build of NSS, however it is broken for other
reasons. It is trying to call "exec" when signing the softoken dll, even without
the patch. exec doesn't exist on my OS/2 system.

Comment 10

[Mike Kaply [:mkaply]]

exec in a makefile is not supposed to invoke exec on a client.

What version of gmake are you using?

Comment 11

[Julien Pierre]

I am using gmake 3.76.1 .
FYI, the problem can be fixed in mozilla/security/nss/cmd/shlibsign/Makefile by
replacing "-exec sign.cmd" with "cmd.exe /c sign.cmd" .

Comment 12

[jhp (no longer active)]

Julien: You should update to make 3.79
(http://www.unixos2.org/sw/pub/binary/make/make-3_79_2a1-r2-bin.zip)

Comment 13

[jhp (no longer active)]

Relyea:  Can we get this checked in for 1.4a?  Thanks...

Comment 14

[Wan-Teh Chang]

Comment on attachment 118337 [details] [diff] [review]
patch v1.3

r=wtc.

Requesting mozilla 1.4a approval.  These are OS/2 build
changes.  All the changes are either in OS/2-only files
or ifdef's with OS/2.

Comment 15

[Asa Dotzler [:asa]]

Comment on attachment 118337 [details] [diff] [review]
patch v1.3

a=asa (on behalf of drivers) for checkin to 1.4a

Comment 16

[Wan-Teh Chang]

Comment on attachment 118337 [details] [diff] [review]
patch v1.3

Javier,

Two questions:

1. In security/coreconf/OS2.mk

>+ifeq (,$EMXOMFLD_LINKER)  # using LINK386.EXE

Can I add parentheses "()" around EMXOMFLD_LINKER?  That is,

>+ifeq (,$(EMXOMFLD_LINKER))  # using LINK386.EXE

2. In security/nss/cmd/shlibsign/sign.cmd

>+'set BEGINLIBPATH=%BEGINLIBPATH%;'dist'\lib'

Can I move 'dist'\lib to the front of BEGINLIBPATH?  That is,

>+'set BEGINLIBPATH='dist'\lib;%BEGINLIBPATH%'

Please let me know if these two suggested changes are fine.

Comment 17

[jhp (no longer active)]

wtc: Yes, both of those changes should be fine.  Thanks for trying to get this in.

Comment 18

[Wan-Teh Chang]

Patch checked in (with the two changes described in comment 16)
on the NSS TIP (NSS 3.8) and NSS_CLIENT_TAG (mozilla 1.4alpha).

Comment 19

[Wan-Teh Chang]

Attachment: Actual checkin

This is Javier's patch v1.3 plus the two changes I suggested
in comment 16.