This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Remove some obfuscation from ${arch}_skip_prologue functions
- From: "Maciej W. Rozycki" <macro at codesourcery dot com>
- To: Doug Evans <dje at google dot com>
- Cc: <gdb-patches at sourceware dot org>, <green at moxielogic dot com>, Yao Qi <yao at codesourcery dot com>
- Date: Thu, 4 Sep 2014 18:51:28 +0100
- Subject: Re: [PATCH] Remove some obfuscation from ${arch}_skip_prologue functions
- Authentication-results: sourceware.org; auth=none
- References: <yjt2y4u1pokj dot fsf at ruffy dot mtv dot corp dot google dot com>
On Wed, 3 Sep 2014, Doug Evans wrote:
> Can I enlist any of the other arch maintainers to regression test this
> there: lm32, mips, arm, nios2, tic6x, tilegx, avr, moxie, sh?
Thanks for working on these clean-ups.
I'll try to push this through testing; what makes me curious about is
whether there's a hidden trap here for compressed ISAs (i.e. MIPS16 and
microMIPS code) where the value of the PC is not the same as the memory
address of the instruction being accessed (the LSB or the ISA bit in the
PC is forcefully set to 1). Another matter may be various MIPS16 call
thunks. I have a vague recollection of my previous encounters with this
code, but it was so many years ago I may have troubles bringing back what
it really was, unless I am able to track down some patches.
One issue here is not every possible execution path is covered in
regression testing, some of the issues in places like this only happen in
situations that are difficult to transform into a proper GDB test case
(and e.g. our MIPS16 thunk tests remain non-functional with stock GCC
because of https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53276).
Maciej