Although still lots need to be done, this is another step towards ARM64/Linux
support for llgs. Note that the list of visible registers is the same as in
RegisterInfos_arm64.h which recently was extracted from Darwin implementation.
For some reason, Darwin implementation authors didn't do
Maybe this looks better?
http://reviews.llvm.org/D5374
Files:
source/Interpreter/CommandObject.cpp
Index: source/Interpreter/CommandObject.cpp
===
--- source/Interpreter/CommandObject.cpp
+++ source/Interpreter/CommandObject.cpp
changed as requested
http://reviews.llvm.org/D5374
Files:
source/Interpreter/CommandObject.cpp
Index: source/Interpreter/CommandObject.cpp
===
--- source/Interpreter/CommandObject.cpp
+++ source/Interpreter/CommandObject.cpp
@@
Many times during my recent activities I saw invalid frame error message. I
found it totally useless - it does not give any clue why the error happened. It
turned out that it can occur due to obvious reasons: no target selected, no
process launched etc. With this patch it is easier to guess
This solves problem described in D5232, for llgs this time.
http://reviews.llvm.org/D5341
Files:
source/Plugins/Process/Linux/NativeProcessLinux.cpp
Index: source/Plugins/Process/Linux/NativeProcessLinux.cpp
===
---
Seems like in my comment to this patch I went too far with the last
statement on llgs - I was refering to the code I'm working on and for a
moment I've forgotten that it's something not complete and not published
yet.
http://reviews.llvm.org/D5232
Apparently, PEEKUSER/POKEUSER is something x86 specific, so I had to rework it
for AArch64. This fixes assertion that occurs whenever lldb started on AArch64
device tried to read PC register (or any other register) - this did not occur
in llgs which already does it similar way.
Corrected version, overhead 'if (success)' removed - it consisted of two if's,
first was anded with (reg_info-byte_offset 0x1) which can never be true on
ARM64, second 'if' can never be true because first 'if' action would never
happen.
Note that my RegisterContextPOSIXProcessMonitor_arm64.cpp
Hi Todd,
I'm sorry for late response, I'm doing things on the run for last three
days, hopefully things will settle down soon. As for my patches I can see
that most of them are already commited. The ones that are waiting I listed
below:
http://reviews.llvm.org/D4579 - accepted, not commited
Hi Todd,
I'm sorry for late response, I'm doing things on the run for last three
days, hopefully things will settle down soon. As for my patches I can see
that most of them are already commited. The ones that are waiting I listed
below:
http://reviews.llvm.org/D4579 - accepted, not commited
More on registering context on arm64. Depends on D4580 which although accepted,
does not seem to be commited yet (I guess due to Host vs HostInfo problem).
http://reviews.llvm.org/D5089
Files:
source/Plugins/Process/POSIX/CMakeLists.txt
source/Plugins/Process/POSIX/POSIXThread.cpp
I didn't notice this one in my previous commit that added breakpoints opcode
for ARM64...
http://reviews.llvm.org/D5078
Files:
source/Plugins/Process/POSIX/ProcessPOSIX.cpp
Index: source/Plugins/Process/POSIX/ProcessPOSIX.cpp
===
The opcode was taken from gdb sources.
break set -n main
Breakpoint 1: where = example`$x, address = 0x00400c60
Note that for me, gdb worked only with statically linked binaries on emulated
AArch64 machine, therefore from now on I'm testing lldb with statically linked
binaries too.
On gdb, the same breakpoint in the same binary is situated 8 bytes further...
(gdb) break main
Breakpoint 1 at 0x400c68
http://reviews.llvm.org/D4969
___
lldb-commits mailing list
lldb-commits@cs.uiuc.edu
D4488 was completely covered by later patch already on LLVM's master.
http://reviews.llvm.org/D4580
___
lldb-commits mailing list
lldb-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits
Similar patch was accepted later which adds missing R_AARCH64_JUMP_SLOT
constant, so it builds fine now.
http://reviews.llvm.org/D4579
___
lldb-commits mailing list
lldb-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits
I really don't like to put brackets around objects/variables/constants - sizeof
requires them only for type names. I'm doing this only to fit into existing
code.
http://reviews.llvm.org/D4969
Files:
source/Plugins/Platform/Linux/PlatformLinux.cpp
I looked how it is done in gdb, and found that this piece of code must be
rewritten.
http://reviews.llvm.org/D4803
Files:
source/Plugins/Process/Linux/NativeProcessLinux.cpp
Index: source/Plugins/Process/Linux/NativeProcessLinux.cpp
Newer version that fits to recent changes.
http://reviews.llvm.org/D4381
Files:
include/lldb/Core/ArchSpec.h
source/Core/ArchSpec.cpp
Index: include/lldb/Core/ArchSpec.h
===
--- include/lldb/Core/ArchSpec.h
+++
Newer version that fits to recent changes.
http://reviews.llvm.org/D4580
Files:
source/Plugins/Process/Linux/NativeThreadLinux.cpp
source/Plugins/Process/POSIX/POSIXThread.cpp
source/Plugins/Process/Utility/CMakeLists.txt
source/Plugins/Process/Utility/RegisterContextDarwin_arm64.cpp
One more small step toward ARM64/AArch64 support.
WARNING! D4488 on LLVM must be committed first.
http://reviews.llvm.org/D4579
Files:
source/Plugins/ObjectFile/ELF/ELFHeader.cpp
Index: source/Plugins/ObjectFile/ELF/ELFHeader.cpp
New improved version posted.
http://reviews.llvm.org/D4381
Files:
include/lldb/Core/ArchSpec.h
source/Core/ArchSpec.cpp
Index: include/lldb/Core/ArchSpec.h
===
--- include/lldb/Core/ArchSpec.h
+++ include/lldb/Core/ArchSpec.h
This fixes following issue:
(lldb) target select 0
error: index 0 is out of range, valid target indexes are 0 - 4294967295
(lldb) target select 1
error: index 1 is out of range, valid target indexes are 0 - 4294967295
Due to unsigned arithmetic operation, we can see above output which is both
Hey Todd,
That's the plan. This patch is a merely step towards this - it just clears
all the confusion caused by this double naming used by different
toolchains.
On Wed, 9 Jul 2014, Todd Fiala wrote:
Hey Paul,
Are you attempting to build debugserver over there on the linaro Linux build?
Hey Todd,
That's the plan. This patch is a merely step towards this - it just clears
all the confusion caused by this double naming used by different
toolchains.
http://reviews.llvm.org/D4379
___
lldb-commits mailing list
lldb-commits@cs.uiuc.edu
RNBDefs.h changed as requested.
http://reviews.llvm.org/D4379
Files:
source/Core/Error.cpp
source/Host/common/Host.cpp
source/Host/macosx/Host.mm
source/Host/macosx/Symbols.cpp
source/Interpreter/CommandInterpreter.cpp
source/Plugins/ObjectFile/Mach-O/ObjectFileMachO.cpp
That's how Linux started on FoundationV8 emulator names itself:
root@genericarmv8:~# uname -m
aarch64
root@genericarmv8:~# uname -a
Linux genericarmv8 3.10.1.0-1-linaro-vexpress64 #1ubuntu1~ci+130718012724 SMP
Thu Jul 18 01:30:58 UTC 2013 aarch64 GNU/Linux
http://reviews.llvm.org/D4381
On Wed, 2 Jul 2014, Todd Fiala wrote:
Whoops no his name is not Paul Paul. Sorry... Cut and paste the full last
name spelling after writing Paul...
Should be:
Change by Paul Osmialowski
No worries, the essence of the commit message is fine. As for me, I hope
my next AArch64-related
Obsoleted by upstream improvements.
http://reviews.llvm.org/D4057
___
lldb-commits mailing list
lldb-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits
to use the Native* classes as a
follow-up pass. That will enable supporting a remote and local debugging
scenario for a new os/arch with a single set of code for both
remote and local debugging rather than writing a separate remote and local path.
On Mon, Jun 23, 2014 at 1:14 PM, Paul Osmialowski
-Linux box?
-Todd
On Tue, Jun 10, 2014 at 12:51 PM, Paul Osmialowski paw...@king.net.pl wrote:
To allow cross-compilation of Linux/ProcessMonitor.cpp, hide PC-related
stuff behind #ifdef's. Note that struct user used in affected code has a field
named i387 which is x86-specific, therefore
, Todd Fiala wrote:
Ah I see.
I'll have to ping you separately on details - I had started doing something
like that for some chip bringup a few months back. Let me know how it goes!
Will have a look at this shortly.
On Wed, Jun 11, 2014 at 1:08 PM, Paul Osmialowski newch...@king.net.pl wrote
These PTRACE-related enum values are not present everywhere; their presence is
denoted by certain preprocessor definitions in sys/ptrace.h that we can check.
http://reviews.llvm.org/D4091
Files:
source/Plugins/Process/Linux/ProcessMonitor.cpp
Index:
To allow cross-compilation of Linux/ProcessMonitor.cpp, hide PC-related stuff
behind #ifdef's. Note that struct user used in affected code has a field named
i387 which is x86-specific, therefore it should be compiled only for PC.
http://reviews.llvm.org/D4092
Files:
34 matches
Mail list logo
