On 4/27/2012 2:57 PM, Artem Ananiev wrote:
On 4/27/2012 2:46 PM, David Holmes wrote:
Could it be that the parallelism of the new build is simply too high for
your system - hence you get memory exhaustion?
Yes, that's possible, because my desktop has 4Gb RAM. Number of CPU
cores is only 2, though, so it's not massively parallel. I'm now
building in 32-bit mode, will check if it helps.
32-bit build also fails, in the same place as 64-bit:
Compiling 4 files in batch BUILD_BREAKITERATOR
Generating BreakIteratorData
Generating BreakIteratorData_th
Installing META-INF/services/javax.script.ScriptEngineFactory
Installing META-INF/services/com.sun.jdi.connect.spi.TransportService
Installing META-INF/services/com.sun.jdi.connect.Connector
Installing META-INF/services/sun.java2d.pipe.RenderingEngine
Installing META-INF/services/sun.net.spi.nameservice.NameServiceDescriptor
Installing META-INF/services/com.sun.tools.attach.spi.AttachProvider
Installing META-INF/services/javax.print.PrintServiceLookup
Installing META-INF/services/javax.print.StreamPrintServiceFactory
Compiling 8169 files in batch BUILD_JDK
c:\work\ws\build-infra-jdk8\jdk\src\share\classes\java\awt\EventDispatchThread.java:123:
warning: non-varargs call of varargs method with inexact argument type
for last parameter;
final Method evaluateMethod =
Class.forName("sun.lwawt.macosx.EventDispatchAccess").getMethod("evaluate",
null);
^
cast to Class<?> for a varargs call
cast to Class<?>[] for a non-varargs call and to suppress this warning
c:\work\ws\build-infra-jdk8\jdk\src\share\classes\java\awt\EventDispatchThread.java:126:
warning: non-varargs call of varargs method with inexact argument type
for last parameter;
return ((Boolean)evaluateMethod.invoke(cond,
null)).booleanValue();
^
cast to Object for a varargs call
cast to Object[] for a non-varargs call and to suppress this warning
Note: Some input files use or override a deprecated API.
Note: Recompile with -Xlint:deprecation for details.
Note: Some input files use unchecked or unsafe operations.
Note: Recompile with -Xlint:unchecked for details.
2 warnings
321 [main] sh 5300 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
321 [main] sh 5300 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
/bin/sh: fork: retry: Resource temporarily unavailable
2 [main] sh 4440 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
2 [main] sh 4440 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
/bin/sh: fork: retry: Resource temporarily unavailable
11 [main] sh 5820 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
11 [main] sh 5820 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
/bin/sh: fork: retry: Resource temporarily unavailable
415 [main] sh 3396 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
/bin/sh: fork: retry: Resource temporarily unavailableit memory for
stack 0x289000(94208), Win32 error 487
12 [main] sh 6008 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
12 [main] sh 6008 child_info_fork::abort: can't commit memory for
stack 0x289000(94208), Win32 error 487
/bin/sh: fork: Resource temporarily unavailable
CompileJavaClasses.gmk:231: recipe for target
`/cygdrive/c/work/ws/build-infra-jdk8/build/windows-ia32-normal-clientANDserver-release/jdk/newrt.jar'
failedmake[
]: ***
[/cygdrive/c/work/ws/build-infra-jdk8/build/windows-ia32-normal-clientANDserver-release/jdk/newrt.jar]
Error 254
makeMakefile:41: recipe for target `all' failed[1]: *** [all] Error 2
/cygdrive/c/work/ws/build-infra-jdk8/common/makefiles/Makefile:121:
recipe for target `jdk' failed
make: *** [jdk] Error 2
Thanks,
Artem
Thanks,
Artem
David
On 27/04/2012 7:56 PM, Artem Ananiev wrote:
Hi, Magnus,
thanks for your feedback. Here is what I have already tried:
1. Disable anti-virus
2. Upgrade sh/bash/make to the latest or to the stable versions
3. Disable DEP for sh.exe/bash.exe
4. Run rebaseall/peflagsall
5. Re-run make
6. Reboot
It still fails. I cannot say for sure if it fails in the same place, but
always in "jdk". All other workspaces are built fine: hotspot,
langtools, etc.
Thanks,
Artem
On 4/27/2012 12:36 PM, Magnus Ihse Bursie wrote:
On 2012-04-27 09:38, Artem Ananiev wrote:
15830981 [main] sh 7892 C:\cygwin\bin\sh.exe: *** fatal error - fork:
can't reserve memory for stack 0x29BCF0 - 0x2A0000, Win32 error
48715830981 [main] sh 7892 C:\cygwin\bin\sh.exe: *** fatal error -
fork: can't reserve memory for stack 0x29BCF0 - 0x2A0000, Win32 error
487
This is a cygwin problem, which pops up every now and then when running
cygwin.
The basic problem is that cygwin tries to emulate posix behaviour on
windows, but the fork semantics is not available and so it has to be
"faked". Sometimes this fails. It seems to fail more often on
Windows 7.
Running make files often causes a lot of forks, so it increases the
likelihood of failure.
Please just try to re-run make and see if that solves your problem.
If it repeats and seems to happen at the same place, let us known. If
that is the case, we might find some specific circumstances that makes
it more likely to happen, and maybe we can modify that slightly.
/Magnus
