[PATCH] Fix PR39246: -Wuninitialized for partially initialized complex

Warning for use of a partially initialized complex is displayed on the wrong line (the one where the partial initialization is done) and is not consistent between target (or even within a given target, for instance between hardfloat and softfloat on arm-none-eabi target). This patch correctly re

RE: [PATCH][2/3] Fix PR54733 Optimize endian independent load/store

Please find attached the new version of this patch addressing all your comments. ChangeLog are now as follows: *** gcc/ChangeLog *** 2014-05-04 Thomas Preud'homme PR tree-optimization/54733 * expr.c (get_inner_reference): Add a parameter to control whether a MEM_REF s

RE: [PATCH][3/4] Fix PR54733 Optimize endian independent load/store

Please find attached the new version of this patch addressing all suggestions given. I splat the patch into 2 to ease the review: part3 only split stmt manipulation out of optimize_bswap into replace_bswap while part4 contains all the changes needed to support load in host endianness (as opposed to

[PATCH][4/4] Fix PR54733 Optimize endian independent load/store

Here is the fourth and last patch in the series. Its purpose is to add support for bitwise operation from a source operand that are equivalent to a load in the target's endianness. ChangeLog are as follows: *** gcc/ChangeLog *** 2014-05-04 Thomas Preud'homme PR tree-optimization/5473

Re: [PATCH GCC]Fix pr60363 by adding backtraced value of phi arg along jump threading path

On 05/04/14 00:39, bin.cheng wrote:> Hi, I updated and rebased the patch against latest trunk. It passes bootstrap and regression test on x86/x86_64. Also pr60363 is fixed on logical_op_short_circuit targets. Is it OK? Since ssa-dom-thread-4.c is fixed now, I also reverted the XFAIL test for

Re: [PATCH] Clean up and extend VRP edge-assertion code

This patch causes a latent test failure in gcc.dg/uninit-pred-9_b.c due to an oversight in the tree-ssa-uninit code for which I will send a fix shortly.

[patch, libgfortran] PR52539 I/O: Wrong result for UTF-8/UCS-4 list-directed and namelist

With the attached patch, I have taken a different approach. I have created two function pointers in the gfc_unit structure. These pointers are set to more specific versions of next_char and push_char that are geared to the type of I/O being performed. This removes the tests for IO type out of th

Re: [wide-int] Handle zero-precision INTEGER_CSTs again

Kenneth Zadeck writes: > Then with a fixed comment, this patch is fine. OK, here's what I committed. Richard Index: gcc/wide-int.cc === --- gcc/wide-int.cc 2014-05-03 07:59:36.274750108 +0100 +++ gcc/wide-int.cc 2014-05-04

[buildrobot][patch] picochip: Fix int -> enum machine_mode fallout

Hi! I propose the following patch to cure this: g++ -c -g -O2 -DIN_GCC -DCROSS_DIRECTORY_STRUCTURE -fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variad

Re: [buildrobot][patch] picochip: Fix int -> enum machine_mode fallout

On 05/04/14 14:37, Jan-Benedict Glaw wrote: Hi! I propose the following patch to cure this: g++ -c -g -O2 -DIN_GCC -DCROSS_DIRECTORY_STRUCTURE -fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Woverloaded

Re: [patch] change specific int128 -> generic intN

> I'm not aware of any reason those macros need to have decimal values. I'd > suggest removing the precomputed table and printing them in hex, which is > easy for values of any precision. Here's an independent change that removes the decimal table and replaces it with generated hex values. I

Re: [PATCH, RS6000] Fix __builtin_{pack,unpack}_longdouble and __builtin_{pack,unpack}_dec128 builtins

On Fri, May 2, 2014 at 7:34 PM, Peter Bergner wrote: > Currently, the gcc.target/powerpc/pack0[23].c test cases are failing on trunk > and the 4.9 and 4.8 branches. The problem is we either fail to register the > builtins leading to undefined reference errors to the __builtin_* symbol or > we ICE

[patch, libfortran] [4.9/4.10 Regression]PR61049 NIST test FM906 fails

Hi all, The attached patch fixes this regression and retains the desired behavior for PR46800. We don't have a test case for PR46800 since it involves user input at a terminal. (It would be good to get such a test case of course) This patch has been regression tested with 4.9 and 4.10 on x86-64 l

RE: [PATCH][4/4] Fix PR54733 Optimize endian independent load/store

I realized there was a memset missing for the statistics of nop OR bitwise expression. Please consider the attached patch instead. ChangeLog unchanged but reproduced here for convenience: *** gcc/ChangeLog *** 2014-05-04 Thomas Preud'homme PR tree-optimization/54733 * tree-ss

Re: [PATCH, RS6000] Fix __builtin_{pack,unpack}_longdouble and __builtin_{pack,unpack}_dec128 builtins

On Sun, 2014-05-04 at 17:39 -0400, David Edelsohn wrote: > On Fri, May 2, 2014 at 7:34 PM, Peter Bergner wrote: > > @@ -2517,6 +2518,7 @@ extern int frame_pointer_needed; > > #define RS6000_BTM_POPCNTD MASK_POPCNTD/* Target supports ISA > > 2.06. */ > > #define RS6000_BTM_CELL

Re: [PATCH] Clean up and extend VRP edge-assertion code

On Sun, May 4, 2014 at 12:20 PM, Patrick Palka wrote: > This patch causes a latent test failure in gcc.dg/uninit-pred-9_b.c > due to an oversight in the tree-ssa-uninit code for which I will send > a fix shortly. Never mind, I spoke too soon.. I haven't been able to solve this properly.

Re: [patch, libfortran] [4.9/4.10 Regression]PR61049 NIST test FM906 fails

On Mon, May 5, 2014 at 4:15 AM, Jerry DeLisle wrote: > Hi all, > > The attached patch fixes this regression and retains the desired behavior for > PR46800. We don't have a test case for PR46800 since it involves user input > at a > terminal. (It would be good to get such a test case of course) >