|
From: Philippe W. <phi...@sk...> - 2011-10-08 15:06:45
|
I have done the changes needed to have vgdb compiling and working on Android (see https://bugs.kde.org/show_bug.cgi?id=283600 ). As I have no android system, I used the Android emulator part of the SDK to test. However, this emulator emulates an armv5 (so not a cortex a8). To obtain a Valgrind on Android, I followed README.android (there is only one typo in this: need to replace nexus_t by nexus_s IIUC). In terms of source patching, I started by applying the patches c2 and c3 from https://bugs.kde.org/show_bug.cgi?id=276897 (without these patches, the trunk does not compile for armv5). With the patches, this gave a dispatcher that could not execute more than one block : infinite loop producing FASTMISS output lines when enabling the scheduler tracing). With some modifications in the dispatcher (not understanding what I was doing :), Valgrind could execute some more blocks (but fails quickly in the system linker when loading the vg preload .so). With the above, Valgrind executes enough blocks that I could do some manual testing of a gdb running on a fedora/x86 and connecting to the Valgrind gdbserver running on Android emulator. At first sight, gdb+Valgrind gdbserver on Android seems ok. If someone has a patch to obtain a working Valgrind on the Android emulator armv5, I am interested :). Otherwise, any help to do some more validation of the patch in bug 283600 on a (real) android system is welcome. Philippe |