enh-google

define "understand"... it's caused by https://github.com/llvm/llvm-project/blob/main/compiler-rt/lib/profile/InstrProfilingFile.c#L529 but i don't understand why that code is doing what it's doing. why not just check for EEXIST on the mkdir()? apparently i asked...

i've fixed the url so it should work for everyone, since it's an AOSP change.

internal change https://googleplex-android-review.git.corp.google.com/c/platform/tools/base/+/23372540 (that link will only work for google employees :-( ) was submitted 2023-06-06 followed by https://googleplex-android-review.git.corp.google.com/c/platform/tools/base/+/23732318 on 2023-06-20. even though (afaict) these _changes_ aren't public yet, they...

> Maybe we can add this to clang driver once we have tested it? oh, definitely --- any time we land something in AOSP (and we feel like it's _really_...

> I'll replace the previous workaround with -munaligned-access is there a reason we haven't moved this into the driver yet? this flag has been in AOSP for a long time...

the llvm change was merged ... is that cherrypicked for our llvm? (and is that cherrypick in the toolchain that we're working on currently for the forthcoming ndk release, or...

(even if we do nothing, i think this will solve itself over time as we remove guard relating to no-longer-supported API levels...)

> (optionally, eventually by default?) using lldb behind the scenes fwiw, ndk-gdb does now default to lldb. slightly older versions had a --lldb option to request lldb.

> We (Andes Technology) are working on this item now. If we done the imlementation, we will contribute the source. great to hear! it looks like we (Android) haven't been...

i think this is probably a dupe of #450 --- i'm guessing you have a glibc new enough to have removed crypt()? `ldd --version` will tell you your glibc version...