Re: [PATCH] kbuild: remove the target in signal traps when interrupted

[Ingo Molnar]

* Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:

> Reported-by: Ingo Molnar <mingo@xxxxxxxxxx>
> Reported-by: Rob Herring <robh@xxxxxxxxxx>
> Signed-off-by: Masahiro Yamada <masahiroy@xxxxxxxxxx>
> ---
> 
> If you are happy to help test this patch, that will be appreciated.
> 
> Without applying this patch,
> 
>     $ make -j<nr-proc> 2>&1 | tee log
> 
> Then, you will see an error reported in [1].
> You may need to repeat it dozen of times to reproduce it.
> The more CPU cores you have, the easier you will get the error.
> 
> Apply this patch, and repeat the same.
> You will no longer see that error (hopefully).
> 
> 
>  scripts/Kbuild.include | 23 ++++++++++++++++++++++-
>  1 file changed, 22 insertions(+), 1 deletion(-)

I've tested your patch on a system with 64 CPUs on the latest upstream 
kernel, and without your patch I was able to quickly reproduce the 
corrupted .cmd files within 4-5 interrupted kernel builds:

  fs/xfs/libxfs/.xfs_alloc.o.cmd:5: *** unterminated call to function 'wildcard': missing ')'.  Stop.

or:

  kernel/time/.tick-broadcast.o.cmd:5: *** unterminated call to function 'wildcard': missing ')'.  Stop.

or:

  lib/.is_single_threaded.o.cmd:5: *** unterminated call to function 'wildcard': missing ')'.  Stop.

These corrupted .cmd files blocked subsequent additive builds perpetually, 
until I manually removed the corrupted .o.cmd file or did an explicit 'make clean'.

With your patch I've yet to see a single failure, after dozens of attempts.

  Tested-by: Ingo Molnar <mingo@xxxxxxxxxx>

Thanks a lot for taking care of this problem!

	Ingo