70445 - Cannot build debug & opt nspr trees under win32

Status: VERIFIED FIXED

(SeaMonkey :: Build Config, defect, P1)

Description

[cls]

A user posted about this problem in the m.builds newsgroup.  The drawback of
autoconf builds is that we cannot build with multiple <config>.OBJ dirs unless
we figure out a way to also place the generated config files (autoconf.mk) into
a <config>.OBJ directory as well.  Rather than go through that as I'd like
<config>.OBJ subdirs to go away once win32 is switched over to use autoconf, I
have a patch that will create a <config> dir within mozilla/nsprpub/ and do an
objdir build from there.

Comment 1

[cls]

Attachment: nsprpub\gmakefile.win ... used to create debug/opt objdirs

Comment 2

[cls]

Attachment: changes to client.mak to use gmakefile.win

Comment 3

[Wan-Teh Chang]

In nsprpub\gmakefile.win, you say:
# Argh.  nmake keeps the cwd from cmd to cmd and gmake does not
# Furthermore, shmsdos doesn't support '&&' so there's a chance the
# 'cd' could fail and configure would be run in the wrong dir

Aren't we requiring that 'make' and 'sh' from the Cygwin tools
be used?  Are you still trying to support Netscape's Win32 port
of gmake and shmsdos?

Comment 4

[cls]

In trying to address the speed concerns of using the cygwin posix layer for each
command executed, we only require the use of ash for running configure.  There
is a slight slowdown from using ash as opposed to shmsdos but nothing as drastic
as what was estimated before (on w2k, win9x might be a different story).  Unless
more problems like this pop up, I don't see a huge problem in supporting the
netscape tools.

Comment 5

[Brian Ryner (not reading)]

wtc- no, we are not using cygwin's make.  We're supporting the updated (3.79.1)
version of gmake natively compiled for Win32 (and hacked to use shmsdos), for
speed.  This is what's in wintools.zip.

Comment 6

[Matthew Wilson]

This change to client.mak has stopped my build working (on WinNT), I have had 
to revert to 3.95.

Comment 7

[Matthew Wilson]

Please ignore my previous comments, I can no longer reproduce this problem. 
Sorry for wasting time.

Comment 8

[Daniel (Leaf) Nunes]

this looks like it's already in the tree... is that right, cls? r=leaf, 
posthumously

Comment 9

[cls]

Yes, it's in the tree and causing the usual amount of havoc.

Comment 10

[Jon Granrose]

verified.