Le 10/03/2012 18:26, André Hentschel a écrit :
---
programs/winedbg/be_arm.c | 10 ++++++----
1 files changed, 6 insertions(+), 4 deletions(-)
diff --git a/programs/winedbg/be_arm.c b/programs/winedbg/be_arm.c
index c39043c..6c1cc05 100644
--- a/programs/winedbg/be_arm.c
+++ b/programs/winedbg/be_arm.c
@@ -205,13 +205,15 @@ static void be_arm_clear_watchpoint(CONTEXT* ctx,
unsigned idx)
static int be_arm_adjust_pc_for_break(CONTEXT* ctx, BOOL way)
{
+ INT step = (ctx->Cpsr& 0x20) ? 2 : 4;
hmm
the way winedbg works might put us in trouble here...
when winedbg is entered with a break exception:
1/ winedbg calls adjust_pc_for_break(..., TRUE) to get the PC to the
real location of the break insn (and not after, as most CPU report in
break exception handler)
2/ user interacts with winedbg
3/ user resumes execution, and winedbg calls again
adjust_pc_for_break(..., FALSE)
but, the ctx passed here is the context used to resume execution... and
in ARM case Cpsr could have been changed by user in step 2/
leading to wrong PC computation in 3/
therefore, if we want to plug that hole, I'd suggest having two context
used in winedbg (the incoming one (never being changed), and the
current/outgoing one), and that we pass the two contexts in winedbg to
do the right thing here
A+
--
Eric Pouech
"The problem with designing something completely foolproof is to underestimate the
ingenuity of a complete idiot." (Douglas Adams)
