Hi, I do not know what causes this error, but on gentoo, when compiling mozilla: I get this output: /bin/sh: ../../../coreconf/nsinstall/Linux2.4_sparc_glibc_PTH_OPT.OBJ/nsinstall: No such file or directory make: *** [sparc64-unknown-linux-gnu/alg1485.o] Error 127 make: Leaving directory `/var/tmp/portage/mozilla-1.0.1-r1/work/mozilla/security/nss/lib/certdb' Sat Oct 12 18:52:09 Local time zone must be set--see zic manual page 2002 cd certhigh; emake -j1 libs make: Entering directory `/var/tmp/portage/mozilla-1.0.1-r1/work/mozilla/security/nss/lib/certhigh' /bin/sh: ../../../coreconf/nsinstall/Linux2.4_sparc_glibc_PTH_OPT.OBJ/nsinstall: No such file or directory make: *** [sparc64-unknown-linux-gnu/certhtml.o] Error 127 make: Leaving directory `/var/tmp/portage/mozilla-1.0.1-r1/work/mozilla/security/nss/lib/certhigh' Sat Oct 12 18:52:09 Local time zone must be set--see zic manual page 2002 cd pk11wrap; emake -j1 libs make: Entering directory `/var/tmp/portage/mozilla-1.0.1-r1/work/mozilla/security/nss/lib/pk11wrap' /bin/sh: ../../../coreconf/nsinstall/Linux2.4_sparc_glibc_PTH_OPT.OBJ/nsinstall: No such file or directory make: *** [sparc64-unknown-linux-gnu/dev3hack.o] Error 127 make: Leaving directory `/var/tmp/portage/mozilla-1.0.1-r1/work/mozilla/security/nss/lib/pk11wrap' Sat Oct 12 18:52:10 Local time zone must be set--see zic manual page 2002 cd cryptohi; emake -j1 libs make: Entering directory `/var/tmp/portage/mozilla-1.0.1-r1/work/mozilla/security/nss/lib/cryptohi' /bin/sh: ../../../coreconf/nsinstall/Linux2.4_sparc_glibc_PTH_OPT.OBJ/nsinstall: No such file or directory ad nauseum. Here's the thing, nsinstall is actually in: ../../../coreconf/nsinstall/sparc64-unknown-linux-gnu/nsinstall How do I tell mozilla this? Or is this an error on an even more fundamental level?
by the way this happens on 1.0 and 1.0.1, sorry :) and I did apply a patch that I found in another bug report on here about the sparc/sparc64 linux platform. the error persists despite that patch...
==> Build Config
Assignee: asa → seawood
Component: Browser-General → Build Config
QA Contact: asa → granrose
coreconf is NSS.
Assignee: seawood → wtc
Component: Build Config → Build
Product: Browser → NSS
QA Contact: granrose → wtc
Version: other → 3.6
Do you have the environment variable PLATFORM set to sparc64-unknown-linux-gnu when you build Mozilla? That would explain this build failure.
Status: UNCONFIRMED → ASSIGNED
Ever confirmed: true
Version: 3.6 → 3.5
I did that and it didn't work :/
that is to say, it failed compilation with the exact same errors...
Actually, I meant that you should not have PLATFORM set in your environment. Could you show me your environment?
construct root # set BASH=/bin/bash BASH_VERSINFO=(="2" ="05a" ="0" ="1" ="release" ="sparc-unknown-linux-gnu") BASH_VERSION='2.05a.0(1)-release' COLUMNS=80 CVS_RSH=ssh DIRSTACK=() EDITOR=/usr/bin/nano EUID=0 GROUPS=() HISTFILE=/root/.bash_history HISTFILESIZE=500 HISTSIZE=500 HOME=/root HOSTNAME=construct HOSTTYPE=sparc IFS=$' \t\n' INFODIR=/usr/share/info:/usr/X11R6/info INPUTRC=/etc/inputrc LESSOPEN='|lesspipe.sh %s' LINES=24 LOGNAME=root MACHTYPE=sparc-unknown-linux-gnu MAIL=/var/mail/root MAILCHECK=60 MANPATH=/usr/share/man:/usr/local/share/man:/usr/X11R6/man OPTERR=1 OPTIND=1 OSTYPE=linux-gnu PAGER=/usr/bin/less PATH=/bin:/sbin:/usr/bin:/usr/sbin:/usr/local/bin:/opt/bin:/usr/X11R6/bin PIPESTATUS=(="0") PPID=24744 PS1='\[\033[01;31m\]\h \[\033[01;34m\]\W \$ \[\033[00m\]' PS2='> ' PS4='+ ' PWD=/root SHELL=/bin/bash SHELLOPTS=braceexpand:hashall:histexpand:monitor:history:interactive-comments:emacs SHLVL=1 SSH_CLIENT='10.0.0.2 34737 22' SSH_TTY=/dev/pty/s1 TERM=rxvt UID=0 USER=root VIMRUNTIME=/usr/share/vim/vim61 XINITRC=/etc/X11/xinit/xinitrc _=INPUTRC
Your environment looks fine. I have no idea what caused the build error. Since I don't have a Gentoo Linux machine, I won't be able to help much other than pointing you at things to look at. The key to debugging this is to find out which environment or make variable has the value sparc64-unknown-linux-gnu. Note that nothing in your environment has that value. The closest thing in your environment is sparc-unknown-linux-gnu, which doesn't have '64' after 'sparc'. Moreover, it's MACHTYPE that has the value, and our build system does not check the value of MACHTYPE. One thing that would have led to this error is that you set PLATFORM to sparc64-unknown-linux-gnu in your environment. But you didn't do that. So I am at a loss as to what's wrong.
ok so it turns out that gentoo's emerge does, indeed, set the PLATFORM environment variable. The funny thing is that it does this for x86, ppc, and alpha as well, but only affects sparc64 (sparc as well?)...
ok, it was our fault on gentoo for setting the PLATFORM in the system profile (it differed from CHOST). this means mozilla can't be cross-compiled I assume?
Status: ASSIGNED → RESOLVED
Last Resolved: 16 years ago
Resolution: --- → INVALID
Actually this is a bug in our build system. Others have run into this problem before, which is why I asked you if PLATFORM was set in your environment. It is easy to fix this bug. I will attach a patch. It is also easy to work around this bug -- just make sure that PLATFORM is not set in the environment.
Status: RESOLVED → REOPENED
Priority: -- → P1
Resolution: INVALID → ---
Target Milestone: --- → 3.7
yeah, unsetting the PLATFORM variable was what I wound up doing on the sparc64 profiles in Gentoo, and it worked like a charm...would love to see your patch though.
Created attachment 103948 [details] [diff] [review] Proposed patch This bug was introduced in NSS 3.4 when the definition of PLATFORM changed from (in platform.mk) PLATFORM = $(OBJDIR_NAME) to (in config.mk) ifndef PLATFORM PLATFORM = $(OBJDIR_NAME) endif If PLATFORM is already set (for example, as an environment variable), we won't set PLATFORM to the right value. This patch does two things. 1. Do not use the value of PLATFORM from the environment. 2. Remove the obsolete file mozilla/security/coreconf/platform.mk.
Status: REOPENED → ASSIGNED
Summary: Compile failure → The PLATFORM environment variable causes compile failure
Version: 3.5 → 3.4
I checked in the patch. If it is useful to override PLATFORM in our own makefiles (such as the manifest.mn file), we should rename the variable (such as NS_PLATFORM or CORE_PLATFORM) because PLATFORM is a common environment variable. (This is at least the third report of this bug.) I doubt it will be easy to override PLATFORM though because our makefiles inadvertently depend on certain relations between PLATFORM, OBJDIR_NAME, and OBJDIR, so you'd probably need to override all three. For the record, the relations are described below. coreconf/config.mk (the old version has the two lines marked with '-'): -ifndef PLATFORM PLATFORM = $(OBJDIR_NAME) -endif coreconf/location.mk: ifdef BUILD_TREE ifdef LIBRARY_NAME ... OBJDIR = $(BUILD_TREE)/nss/$(LIBRARY_NAME) ... else ... OBJDIR = $(BUILD_TREE)/nss ... endif else ... OBJDIR = $(PLATFORM) ... endif coreconf/UNIX.mk: ifdef BUILD_TREE NSINSTALL_DIR = $(BUILD_TREE)/nss NSINSTALL = $(BUILD_TREE)/nss/nsinstall else NSINSTALL_DIR = $(CORE_DEPTH)/coreconf/nsinstall NSINSTALL = $(NSINSTALL_DIR)/$(OBJDIR_NAME)/nsinstall endif Note that the last two inadvertently imply that PLATFORM is equal to OBJDIR_NAME if BUILD_TREE is not defined. My fix ensures this relation by not allowing PLATFORM to be overridden in coreconf/config.mk, but we could also fix this by using $(PLATFORM) instead of $(OBJDIR_NAME) in the definition of NSINSTALL in coreconf/UNIX.mk, to be consistent with the definition of OBJDIR in coreconf/location.mk. I think the latter is something we should do anyway.
Status: ASSIGNED → RESOLVED
Last Resolved: 16 years ago → 16 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.