Re: [PATCH 0/7] Resume breaking ABI

On 10/5/12 5:21 PM, Keith Packard wrote: Colin Harrison colin.harri...@virgin.net writes: Hi, Thinking about this after strong coffee: all those bit-fields would be better unsigned... Yeah, I agree -- Adam, what do you think? Sounds fine, whatever pacifies the compiler. - ajax

Re: [PATCH 0/7] Resume breaking ABI

Colin Harrison colin.harri...@virgin.net writes: Hi, In xserver/include/dixstruct.h struct _Client: isn't clientState type ClientState not int? I can't figure out if C99 permits enums to be used as bitfield types or not. Anyone got a language lawyer around who can figure this out and explain

Re: [PATCH 0/7] Resume breaking ABI

Hi, Keith, I didn't check the bit packing on this but my compiler gave a scattering of... warning: case label value exceeds maximum value for type which may be best not to ignore. So patched... --- ./include/save_dixstruct.h 2012-10-04 19:31:00.451081236 +0100 +++ ./include/dixstruct.h

Re: [PATCH 0/7] Resume breaking ABI

Colin Harrison colin.harri...@virgin.net writes: Hi, Keith, I didn't check the bit packing on this but my compiler gave a scattering of... Probably needs to be 'unsigned int' instead. I still don't know if we can use the enum type here. -- keith.pack...@intel.com pgpWKe1D8wSkh.pgp

Re: [PATCH 0/7] Resume breaking ABI

Hi, Probably needs to be 'unsigned int' instead. I still don't know if we can use the enum type here. My compiler is happy with 'unsigned int', which is the safest bet I reckon for all (I had to dig out my KR ANSI C on this...and read it once more!) Thanks, Colin Harrison

Re: [PATCH 0/7] Resume breaking ABI

Colin Harrison colin.harri...@virgin.net writes: Hi, Probably needs to be 'unsigned int' instead. I still don't know if we can use the enum type here. My compiler is happy with 'unsigned int', which is the safest bet I reckon for all (I had to dig out my KR ANSI C on this...and read it once

Re: [PATCH 0/7] Resume breaking ABI

Hi, Care to submit a patch which updates all of the changed bitfields to unsigned? Ooops...I meant probably OK for all compilers...not all fields. KR ANSI quotes stuff like... The ANSI standard makes fields even more implementation-dependent then did the first edition. I'm cross-compiling

Re: [PATCH 0/7] Resume breaking ABI

Hi, Thinking about this after strong coffee: all those bit-fields would be better unsigned... --- ./include/save_dixstruct.h 2012-10-05 01:24:03.981982326 +0100 +++ ./include/dixstruct.h 2012-10-05 20:13:19.803253520 +0100 @@ -90,12 +90,12 @@ typedef struct _Client { Mask

Re: [PATCH 0/7] Resume breaking ABI

Colin Harrison colin.harri...@virgin.net writes: Hi, Thinking about this after strong coffee: all those bit-fields would be better unsigned... Yeah, I agree -- Adam, what do you think? --- ./include/save_dixstruct.h 2012-10-05 01:24:03.981982326 +0100 +++ ./include/dixstruct.h

Re: [PATCH 0/7] Resume breaking ABI

Adam Jackson a...@redhat.com writes: This series is also available as the 'server-1.4-abi-churn' branch of: git://people.freedesktop.org/~ajax/xserver.git Merged. 0b02150..4dd5989 master - master -- keith.pack...@intel.com pgptWqJoJHLlI.pgp Description: PGP signature

Re: [PATCH 0/7] Resume breaking ABI

Hi, In xserver/include/dixstruct.h struct _Client: isn't clientState type ClientState not int? Thanks, Colin Harrison ___ xorg-devel@lists.x.org: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info:

[PATCH 0/7] Resume breaking ABI

Mostly a resend of a now familiar series, but with two minor changes. One is just bumping the ABI version. #4 in the series is a slightly more aggressive packing than I'd sent before, but it gets ClientRec inside of two cachelines. Which is nice. This series is also available as the