On 9/23/25 14:56, Alyssa Ross wrote:
> Demi Marie Obenour <demiobenour@gmail.com> writes:
> 
>> On 9/23/25 10:35, Alyssa Ross wrote:
>>> Demi Marie Obenour <demiobenour@gmail.com> writes:
>>>
>>>> On 9/21/25 14:36, Alyssa Ross wrote:
>>>>> Demi Marie Obenour <demiobenour@gmail.com> writes:
>>>>>
>>>>>> Don't call it if the target directory already exists.
>>>>>>
>>>>>> Signed-off-by: Demi Marie Obenour <demiobenour@gmail.com>
>>>>>> ---
>>>>>>  scripts/make-erofs.sh | 11 ++++++++---
>>>>>>  1 file changed, 8 insertions(+), 3 deletions(-)
>>>>>>
>>>>>> diff --git a/scripts/make-erofs.sh b/scripts/make-erofs.sh
>>>>>> index ac62a65f53e0b6417b164f475a537960fc5203bc..30b83b0b41cbe7bc4fd7786cfcdddcf10c78cc5a 100755
>>>>>> --- a/scripts/make-erofs.sh
>>>>>> +++ b/scripts/make-erofs.sh
>>>>>> @@ -44,10 +44,15 @@ while read -r arg1; do
>>>>>>  	# a subset of all paths, but this subset includes all of
>>>>>>  	# the ones passed in practice other than /.
>>>>>>  	case $arg2 in
>>>>>> -	(*/*) parent=${arg2%/*};;
>>>>>> -	(*) parent=.;;
>>>>>> +	(*/*)
>>>>>> +		# Make the parent directory if needed
>>>>>> +		parent=$root/${arg2%/*}
>>>>>> +		if [ ! -d "$parent" ]; then mkdir -p -- "$parent"; fi
>>>>>> +		;;
>>>>>> +	(*)
>>>>>> +		# Parent $root which definitely exists
>>>>>> +		;;
>>>>>>  	esac
>>>>>> -	mkdir -p -- "$root/$parent"
>>>>>>  	cp -RT -- "$arg1" "$root/$arg2"
>>>>>>  done
>>>>>
>>>>> I saw a statistically significant speedup in hyperfine. :)
>>>>>
>>>>> Reviewed-by: Alyssa Ross <hi@alyssa.is>
>>>>
>>>> Yay!  This is independent of the rest of the series, so
>>>> feel free to commit.
> 
> It actually doesn't seem to be.  I get a "Permission denied" build
> failure when I apply it on top of 0fcf508 ("scripts/make-erofs.sh: Avoid
> calls to dirname"), but I think I didn't see that failure when I applied
> patch 1 as well.  We can either just keep this in the queue until we've
> figured out what to do about patch 1, or you can investigate and
> resubmit a standalone version of this patch.  Up to you.

I think it is best to include patch 1.  It speeds things up, and
I consider the complexity to not be a serious problem.  The code
that generates the file lists already guarantees this.
-- 
Sincerely,
Demi Marie Obenour (she/her/hers)