[OE-core] [PATCH] busybox: Guard against interrupted compiles

[Mario Domenech Goulart]

On Mon, 23 Jan 2017 13:05:48 +0000 Richard Purdie <richard.purdie at linuxfoundation.org> wrote:

> On Mon, 2017-01-23 at 13:56 +0100, Mario Domenech Goulart wrote:
>> 
>> On Mon, 23 Jan 2017 12:35:52 +0000 Richard Purdie <richard.purdie at linuxfoundation.org> wrote:
>>
>> > On Mon, 2017-01-23 at 12:26 +0000, Richard Purdie wrote:
>> > > 
>> > > diff --git a/meta/recipes-core/busybox/busybox.inc
>> > > b/meta/recipes-
>> > > core/busybox/busybox.inc
>> > > index 1f4a48c..f247e8d 100644
>> > > --- a/meta/recipes-core/busybox/busybox.inc
>> > > +++ b/meta/recipes-core/busybox/busybox.inc
>> > > @@ -141,6 +141,10 @@ do_compile() {
>> > >  	unset CFLAGS CPPFLAGS CXXFLAGS LDFLAGS
>> > >  	if [ "${BUSYBOX_SPLIT_SUID}" = "1" -a x`grep
>> > > "CONFIG_FEATURE_INDIVIDUAL=y" .config` = x ]; then
>> > >  	# split the .config into two parts, and make two busybox
>> > > binaries
>> > > +		if [ -e .config.org ]; then
>> > > +			# Need to guard again an interrupted
>> > > do_compile - restore any backup
>> > > +			cp .config.orig .config
>> > > +		fi
>> > I have fixed the typo...
>> Wouldn't it be better to have something like
>> 
>>    cp .config.orig .config || true
>> 
>> instead, to prevent race conditions?
>
> Prevent which race condition? I'd hope at this point we're the only
> thing touching the .config file?

Yeah, I hope so.

I mean, the "if exists(file){ do_something_with(file) }" idiom is
intrinsically subject to race conditions (time of check to time of use).
If we can [easily] avoid them, why not?

All the best.
Mario
-- 
http://parenteses.org/mario