[Toybox] getprop, setprop, NDK, git/gerrit woes

[enh]

On Sat, Apr 18, 2015 at 6:14 PM, Rob Landley <rob at landley.net> wrote:
> On Sat, Apr 18, 2015 at 6:24 PM, enh <enh at google.com> wrote:
>> i implemented getprop and setprop and switched over internally:
>>
>> commit 2e4c96533c1b7ffb5f303c9589a6b4a1de6be560
>> Author: Elliott Hughes <enh at google.com>
>> Date:   Tue Apr 7 21:41:32 2015 -0700
>>
>>     Implement getprop for toybox.
>>
>>     Change-Id: I10cff49b7c09e9d93b3d53848f2e7d590a045405
>>
>> commit 629526f72ccd2853eadfc1905b6f9f45bc2f6300
>> Author: Elliott Hughes <enh at google.com>
>> Date:   Tue Apr 7 15:46:00 2015 -0700
>>
>>     Implement setprop for toybox.
>>
>>     Because bionic/libcutils doesn't wait for init, I've duplicated a lot
>>     of the input validation that init does so we can give good error messages.
>>
>>     Change-Id: I175421f7409f88813b88109f58825ce5649f1550
>>
>> the problem with these changes is that the properties API isn't public
>> API. so not only do i have an #ifdef so i can build on the host with
>> the same .config file (because having the host is a nice sanity check
>> for all the other commands), you can't actually build this with the
>> NDK either. so maybe you don't even want these patches since they're
>> only really useful in an Android platform tree? dunno. that's your
>> decision. (and the reason why i didn't bother mentioning these patches
>> already. i'm not hiding them or anything :-) )
>
> I need the occasional reminder about stuff you've already mentioned
> anyway, because my todo heap forms archaological layers and then
> compost.
>
>> i've done a couple of git fetches and merges from you since you
>> switched and the one that includes changes i'd already committed
>> confused gerrit. it didn't like seeing changes with new SHAs (because
>> you'd cherrypicked) but the same Change-Id (which is a gerrit thing in
>> the commit message). gerrit won't let me _not_ use a Change-Id (in
>> fact, i never do --- they're always added by gerrit for me). i'm not
>> sure what the best way to avoid this is. the git folks had me do this:
>>
>>   git push aosp HEAD:refs/for/master%base=$(git rev-parse
>> HEAD^),base=$(git rev-parse HEAD^2)
>
> *blink* *blink*
>
> Ok?

i'm glad it's not just me who has no idea what that means!

>> which at least lets me upload to gerrit, but then i hit a gerrit bug
>> that means i can't review the changes there anyway, so i may as well
>> have just pushed directly, avoiding gerrit :-(
>
> If you figure out how I can make less work for you, let me know. I am
> _so_ not a git expert, having avoided doing more than absolutely
> necessary with it (basically using the linux-kernel tree in a read
> only manner) for aesthetic reasons for many years. The entire switch
> to git was to make life easier for the contributors (I was happy with
> mercurial).
>
> How about if I grab the getprop/setprop commits you did but edit _out_
> the change ID from the commit message?

i only play a doctor on tv --- i never venture off the most-traveled
git roads :-) but, yeah, i think if cherrypicking is easiest for you
(and i think it is, because otherwise you have to deal with our
checked-in files [at least until i get time to look at doing that as
part of the build, directly into out/]) then removing the Change-Id
line is the only thing i can think of. and there's something to be
said for low-tech.

like i said, i'll keep sending non-Android stuff here first anyway, so
hopefully this shouldn't happen too often.

 -e

 1429418848.0