The webapprt-test-chrome test runs on the Linux ASAN builds all fail on "AddressSanitizer: global-buffer-overflow": 19:08:27 INFO - ==1794==ERROR: AddressSanitizer: global-buffer-overflow on address 0x7fbf1c1d9e28 at pc 0x7fbf09d0b050 bp 0x7fffe82fae30 sp 0x7fffe82fae28 19:08:27 INFO - READ of size 2 at 0x7fbf1c1d9e28 thread T0 19:08:28 INFO - #0 0x7fbf09d0b04f in InitTypeClasses /builds/slave/ced-l64-asan-00000000000000000/build/source/js/src/ctypes/typedefs.h:75 19:08:28 INFO - #1 0x7fbf09d0b04f in JS_InitCTypesClass(JSContext*, JS::Handle<JSObject*>) /builds/slave/ced-l64-asan-00000000000000000/build/source/js/src/ctypes/CTypes.cpp:1334 19:08:28 INFO - #2 0x7fbf08ba2973 in InitAndSealCTypesClass /builds/slave/ced-l64-asan-00000000000000000/build/source/toolkit/components/ctypes/ctypes.cpp:91 … 19:08:28 INFO - 0x7fbf1c1d9e28 is located 16 bytes to the right of global variable 'ffi_type_sint8' from '/builds/slave/ced-l64-asan-00000000000000000/build/source/js/src/ctypes/libffi/src/types.c' (0x7fbf1c1d9e00) of size 24 19:08:28 INFO - 0x7fbf1c1d9e28 is located 8 bytes inside of global variable 'ffi_type_uint8' from '/builds/slave/ced-l64-asan-00000000000000000/build/source/js/src/ctypes/libffi/src/types.c' (0x7fbf1c1d9e20) of size 24 19:08:28 INFO - 0x7fbf1c1d9e28 is located 24 bytes to the left of global variable 'ffi_type_void' from '/builds/slave/ced-l64-asan-00000000000000000/build/source/js/src/ctypes/libffi/src/types.c' (0x7fbf1c1d9e40) of size 24 19:08:28 INFO - SUMMARY: AddressSanitizer: global-buffer-overflow /builds/slave/ced-l64-asan-00000000000000000/build/source/js/src/ctypes/typedefs.h:75 InitTypeClasses - https://tbpl.mozilla.org/php/getParsedLog.php?id=42813301&tree=Cedar See the attachment for the full stack. More examples: https://tbpl.mozilla.org/php/getParsedLog.php?id=42410829&tree=Cedar https://tbpl.mozilla.org/php/getParsedLog.php?id=42303147&tree=Cedar

Probably the same problem we run into with toolkit/webapps/ tests that execute the app: bug 989569.

Yes, the stack dump is the same.

Looks like the problem is in js-ctypes, in particular I think it comes from osfile's usage of js-ctypes (because looks like it's the only module that uses js-ctypes on Linux).

We're really close to being able to run runtime tests automatically. This is the only remaining blocker.

Here's a recent log: https://tbpl.mozilla.org/php/getParsedLog.php?id=51164486&tree=Cedar. Yoric: any chance you have an idea why this is happening and how to go about fixing it?

I took a quick look at it, and I'm afraid I have no clue. The stack trace is not very convincing, too. I have little to no experience with ASAN, but would a gdb/lldb run be able to pinpoint exactly when the overflow takes place?

The web runtime has been removed in bug 1238079.