https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #20 from Sergey Anikin ---
(In reply to comment #19)
> Yes, I've built Mesa 9.0.2 / LLVM 3.2 with the compiler hotfix and it works.
> The test application also works.
Thanks a lot for the confirmation, Keith!
--
You are receiving t
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #19 from Keith Kriewall ---
Yes, I've built Mesa 9.0.2 / LLVM 3.2 with the compiler hotfix and it works.
The test application also works.
--
You are receiving this mail because:
You are the assignee for the bug.
___
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #18 from Sergey Anikin ---
(In reply to comment #17)
> I posted this issue to Microsoft and they have released a hotfix for Visual
> Studio 2010 SP1 to correct it:
> http://support.microsoft.com/kb/2836024
As soon as you try it (an
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #17 from Keith Kriewall ---
I posted this issue to Microsoft and they have released a hotfix for Visual
Studio 2010 SP1 to correct it:
http://support.microsoft.com/kb/2836024
--
You are receiving this mail because:
You are the ass
https://bugs.freedesktop.org/show_bug.cgi?id=58718
José Fonseca changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|---
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #15 from José Fonseca ---
(In reply to comment #14)
> I posted a series of patches to mesa3d-dev which seems to fix the "inline"
> issue.
I pushed these now, the most important being
commit 57cd1d1454653f778837eec0ee5d4060bc59c5ba
A
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #14 from José Fonseca ---
(In reply to comment #13)
> VS 2012 refuses to compile Mesa due to macro definition of "inline".
>
> I've tried downloading 9.0.3 and 9.1, no change. I had expected to see
> inline/INLINE changes in the sou
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #13 from Keith Kriewall ---
VS 2012 refuses to compile Mesa due to macro definition of "inline".
F:\Program Files (x86)\Microsoft Visual Studio
11.0\VC\INCLUDE\xkeycheck.h(199): warning C4005: 'inline' : macro redefinition
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #12 from José Fonseca ---
It looks like Visual Studio 2012 (must be the final version, and not the
earlier preview) compiles the fd58718.zip test case correctly. I don't know if
it fully fixes all bit field usage in Mesa, or just thi
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #11 from Keith Kriewall ---
I just tried that (signed fields ahead of unsigned) and it didn't help in this
case. The modified struct began as:
struct prog_src_register
{
GLint Index:(INST_INDEX_BITS+1); /**< Extra bit here for si
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #10 from Brian Paul ---
Can you try reordering the fields such that all the signed fields come before
the unsigned fields (or vice versa)? I was tinkering with a test program and
that seemed to help, but I haven't done extensive test
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #9 from Keith Kriewall ---
Sorry, I didn't mean to imply that the signed issue is causing this problem.
I've tried increasing the 'File' bit field size by one, and it made no obvious
difference. I just wanted to note the difference
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #8 from Roland Scheidegger ---
(In reply to comment #7)
> In case it helps, it appears that MSVC always treats enum values as signed
> int. E.g. see:
>
> http://compgroups.net/comp.lang.c++/problem-with-visual-c++-7.1.3088-and-bit-
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #7 from Keith Kriewall ---
In case it helps, it appears that MSVC always treats enum values as signed int.
E.g. see:
http://compgroups.net/comp.lang.c++/problem-with-visual-c++-7.1.3088-and-bit-fields/1013665
GCC appears to use un
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #6 from José Fonseca ---
I've been fiddling with this, and I only see two things that workaround this
MSVC bug:
- disable all optmizations (/O0 isntead of /O2 or /O1)
- force all bit fields from struct prog_src_register to have the sa
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #5 from José Fonseca ---
Created attachment 76244
--> https://bugs.freedesktop.org/attachment.cgi?id=76244&action=edit
fdo58718.zip
It looks like MSVC is generating invalid code somehow. I've attached a
self-contained test program
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #4 from José Fonseca ---
Created attachment 76199
--> https://bugs.freedesktop.org/attachment.cgi?id=76199&action=edit
framebuffer2.trace
The attached trace allows me reproduce the issue with a release/profile MSVC
build of mesa.
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #3 from Keith Kriewall ---
Neither, I'm building both with Visual Studio 2010. LLVM has a
project/solution as generated by CMake, and Mesa builds via scons in a Command
Prompt environment configured as follows:
F:\proj\Mesa-9.0.
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #2 from Brian Paul ---
Using gcc or clang? Which version?
--
You are receiving this mail because:
You are the assignee for the bug.
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.or
https://bugs.freedesktop.org/show_bug.cgi?id=58718
Keith Kriewall changed:
What|Removed |Added
CC||keith.kriewall@attachmate.c
https://bugs.freedesktop.org/show_bug.cgi?id=58718
--- Comment #1 from Keith Kriewall ---
I've run into this with Mesa 9.0.2 and LLVM 3.2. It appears that the bitfields
in the prog_src_register struct (/src/mesa/program/prog_instruction.h) do not
translate correctly to the src_register() paramet
https://bugs.freedesktop.org/show_bug.cgi?id=58718
Sergey Anikin changed:
What|Removed |Added
See Also||https://bugs.freedesktop.or
22 matches
Mail list logo
