Closed muhlpachr closed 5 years ago
You mean the latest version of Windows Build Tools crashes on the latest version of Win10 64-bits?
Yes latest version of Windows Build Tools crashes on the latest version of Win10 64-bits but on some instances only. I am not able to distinguish why on some instances fails and on other works fine. I am now on MSW10 instance where latest BusyBox from Windows Build Tools crashes and latest BusyBox from Scoop works fine.
Are you using the 64-bits of the Windows Build Tools?
Yes, 64-bits of the Windows Build Tools. Behaves same as 32-bits version. But crashes are related to certain use only, in this case with make. Executing just sh.exe seams fine.
Are the make files generated by Eclipse?
No, ordinary make files written by hand.
Hmmm... I'm afraid I'll not be able to reproduce the problem.
Any idea if BusyBox from Scoop is newer?
If so, I can try to update the Windows Build Tools to the latest BusyBox.
Yes, it is hard to reproduce. I have alike issues with Scoop version of BusyBox with ls - it sometimes crashes. Scoop version is usually newer, but update to newer version will not solve root issue probably because I am having alike issues with various versions of BusyBox.
BusyBox might be too light for your use case. Did you try the sh
binary provided by the Git for Windows package?
Well it works fine in Linux environment, so I am sure it is heavy enough ;-) There is BusyBox version od Git for Windows but it crashes more often. Heavier sh from Git for Windows has issues with paths who contain spaces unfortunately.
Heavier sh from Git for Windows has issues with paths who contain spaces unfortunately.
generally make
has problems with names which contain spaces, I don't know of such problems with sh
.
I updated the build script, it is functional now.
if you want to build your own make you can start from the current configuration.
Thank you. On some MSW10 works fine, on some binaries crashes sometimes. I am not able to distinguish configuration differences unfortunately.
probably your make files use some feature that requires a shell command/variable/environment/etc available only on some machines.
Even if that is a case, BusyBox should not crash with unhandled exception, is not ? It should just report error message then.
right, BusyBox should not crash.
but until you can reproduce the problem, I'm afraid even the BusyBox author can do little to help.
Is there any chance to compile it way that faults info provide more information about nature of crash and provide leads to find root source of issue please ?
The build script is functional, you can experiment with any compile options you think useful.
Do you have any hints what options/debug version of libs to use to get useful info in runtime in case of crash please ?
I don't have any experience with BusyBox. I would enable some debug options, to produce a detailed log.
What debug options are you suggesting to catch win32 unhandled exceptions please ? I am not able to recognize any relevant one.
Sorry, my experience with win32 is close to none. Check the BusyBox sources for debugging support, and possibly enable some trace.
I do not understand win32 either. It seams to me there are not any win32 related code in BusyBox but it uses POSIX API emulation layer - there is root issue probably.
I can confirm the random crashes of busybox/sh.exe when using with make. My system is:
Sorry, forgot to mention the version of busybox: $> BusyBox v1.28.0.git (2018-01-03 14:23:53 EST) multi-call binary
Could you try the latest version too? (published a few days ago)
Ok, version $> BusyBox v1.29.0.git (2018-04-28 12:09:03 EDT) multi-call binary fixes this issue on my machine.
Thank you for your work in this very helpful project!
thank you, good luck with your projects!