Managed resource parser needs sorting tweak

[garuma]

Comparing the output of aapt and the managed resource parser with https://github.com/toggl/mobileapp I noticed a difference in the way IDs are assigned:

Basically the gist of it seems to be that aapt assigns its IDs taking case into order (but then outputs it in a case insensitive-fashion).

[dellis1972]

This sounds familiar... I need to look at the managed parser again anyway to try to speed it up, this sounds like a good excuse to start looking :)

[dellis1972]

@garuma so is the problem with aapt2 or aapt? I've tested with aapt2 and it seems to be producing the same results. The Expand item is getting a value of 0x7F020004.

[dellis1972]

if its aapt I'm not sure if we even should support compatibility with it. Since we want to move to aapt2 asap.

[garuma]

🤔 strange it looks like the project does force aapt2 (AndroidUseAapt2 in the csproj is set to true). I can retry with a more recent X.A build to make sure and get a binlog.

[garuma]

binlogs.zip

Attached the binlogs for a design-time build and a full build, they seem to indicate aapt2 is indeed used.

Despite that it still ends up generating the different results that I presented initially. For X.A I'm using the latest build from #4190, my toggl is at commit cfae229995

[dellis1972]

hmm, for some reason I can't decode those binlog files

error MSB4025: The project file could not be loaded. '�', hexadecimal value 0x01, is an invalid character. Line 1, position 1.

I get that for both Toggl.Droid_Debug_AnyCPU_Build.buildlog. I'll try on windows tomorrow.

[dellis1972]

@garuma so with the images you posted. The one on the right are the values I get locally.
What I did was take the intermediate output dir of the Toggl app and cleaned it out so only the res and lp directories with resources were left. The wrote a test to test the parse against that. I'm getting the same values as the image on the right. That lines up with the R.txt that gets generated from aapt2 on the same folder.