Closed Bug 524149 Opened 15 years ago Closed 12 years ago

make -jN frequently hangs in MozillaBuild 1.4

Categories

(Firefox Build System :: MozillaBuild, task)

Product:
Firefox Build System
Component:
MozillaBuild
Platform:
x86
Windows XP
Type:
task
Priority:
Not set
Severity:
normal

Tracking

(Not tracked)

Status:
RESOLVED WONTFIX

People

(Reporter: ted, Unassigned)

Details

I've seen personally (and heard numerous reports) that parallel make hangs in MozillaBuild 1.4. We did take a MSYS update to fix Windows x64 compat, so it's possible that's the root cause.

I've built with 1.3 -j4 on my Windows XP, Core 2 Duo system numerous times without problems, so I know it's possible in some setups.
If we use -j>1, the build seems to progress fine, and then at some (random?) time during the build will hang in the make process. We're hitting this with *every* attempt on a loaner BL460 + 2x72GB 15k HDD. As far as we can tell, there's no error messages in build logs.

If we use -j1, then the same build works fine.

Note: this is with MozillaBuild1.4, *and* also with MozillaBuild1.3.
fyi: we use MozillaBuild1.4 with -j4 on our single-core VMs without any problems. This physical machine is a mulit-core machine.
ted: any progress on this? jhford is hitting this consistently on a new 1u physical box we are trying out. Note: we are running Win2k3, while you are running WinXP.

jhford: can you try with "-j1" to see if those builds at least complete?
I haven't done any testing to try to make this work. If we're going to need parallel builds for the new boxes I'll just finish the few remaining PyMake bugs. I don't think trying to debug msys/gnu make is a good use of anyone's time.
if -j1 is the same as no -j flag, then yes the machines are building fine.  When I tried -j4 i was getting a bunch of hung make processes
The Simple Firefox build instructions at
https://developer.mozilla.org/En/Simple_Firefox_build#Building_Firefox
tell people to create a mozconfig file that passes -j4 to make.
Until this bug is fixed, the build instructions should note that
-j4 should not be used on Windows.
I updated the instructions.

Wan-Teh, you could have done that too, you know, the page isn't locked down or anything.  :-)  (I assume you already have an MDC login, would be surprised if you'd never needed one for modifying NSPR docs or something.)
(In reply to comment #5)
> I haven't done any testing to try to make this work. If we're going to need
> parallel builds for the new boxes I'll just finish the few remaining PyMake
> bugs. I don't think trying to debug msys/gnu make is a good use of anyone's
> time.

Ted, is it fair to say this is WONTFIX as far as our end of it goes; but if it is fixed upstream in some way, we'd be happy to take the upgrade (if it doesn't break anything more important).
FWIW, Win7, 4 core dell, w/-j4, no major issues. Once in a blue moon make will hang randomly during a build, but I've been experiencing that for ages.
FWIW, I got this nearly constantly on David Bolter's Win7 machine.  CCing him to see if he can provide more info.
(In reply to comment #10)
> Ted, is it fair to say this is WONTFIX as far as our end of it goes; but if it
> is fixed upstream in some way, we'd be happy to take the upgrade (if it doesn't
> break anything more important).

I have no intention of spending any time debugging msys or GNU make.
(In reply to comment #12)
> FWIW, I got this nearly constantly on David Bolter's Win7 machine.  CCing him
> to see if he can provide more info.

Parallel builds worked fine on this machine in the past.
FWIW, I'd try using gmake 3.82 now that it is out.
I think this is WONTFIX in favor of moving to pymake by default on Windows (Bug 593585)
I still have a problem with compiling comm_central. I am completly new in mozilla. My mozconfig lokks like that:
# My first mozilla config
mk_add_options MOZ_OBJDIR=C/thunderbird_src/objects
mk_add_options MOZ_MAKE_FLAGS="-j1"
ac_add_options --enable-calendar
ac_add_options --enable-application=mail

Even if i have -j1 the build process hangs here and there.

If i try pymake it always hangs at the same line in the beginning of the build process:
checking host system type...

I have a win32 xp sp3 system with 3GB (actually 2 then, i think) and an intel core2 duo cpu.
Do we care about this now that pymake is the default for Windows?
It's been WONTFIX for a while, I never intended to do anything about this. We don't technically require Pymake on Windows yet, but we ought to.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → WONTFIX
Product: mozilla.org → Firefox Build System
You need to log in before you can comment on or make changes to this bug.