[Bug target/88799] New: Arm -mcpu=PROCESSOR does not result in assembly directives for .arch and .arch_extension

2019-01-10 Thread pcarroll at codesourcery dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88799

Bug ID: 88799
   Summary: Arm -mcpu=PROCESSOR does not result in assembly
directives for .arch and .arch_extension
   Product: gcc
   Version: 8.2.1
Status: UNCONFIRMED
  Severity: normal
  Priority: P3
 Component: target
  Assignee: unassigned at gcc dot gnu.org
  Reporter: pcarroll at codesourcery dot com
  Target Milestone: ---

Created attachment 45410
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=45410&action=edit
Example test case with embedded PLDW instruction

In June 2017, the mechanism for handling -march and -mcpu was modified for the
GCC 8 Arm compiler. In December 2017, the Arm compiler began emitting .arch
directives within each function that was being emitted, if the current
architecture had changed from the previous function.  The problem seems to be
that this new behavior does not track architecture extensions that have been
specified on the command line or in the source.

For example, if I compile a C file with the command line "-S -mcpu=cortex-a9"
and the C file contains an asm instruction that is beyond the basic
architecture instruction set, the GCC 8 compiler generates this assembly code: 

.cpu cortex-a9
.eabi_attribute 20, 1
:
.global myfunc
.arch armv7-a
.syntax unified
.arm
:
pldw L1
L1:
:

When this is assembled, I get the message
Error: Selected processor does not support `pldw L1` in ARM mode

If I insert a '.arch_extension mp' directive after the '.arch armv7-a', the
assembly file does assemble without error.

This is a regression from GCC 7.

[Bug c++/49004] Improve the error message for linking failure

2011-05-15 Thread pcarroll at codesourcery dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49004

Paul Carroll  changed:

   What|Removed |Added

 CC||pcarroll at codesourcery
   ||dot com

--- Comment #2 from Paul Carroll  2011-05-16 
05:34:55 UTC ---
There is at least one way to modify the binutils linker to find the name of the
DSO that contains the hidden symbol reference.  The solution I implemented last
month hasn't been reviewed yet, so I have not gotten any feedback on whether it
is acceptable or not to put into ld.