[Mesa-dev] renderbuffer-cleanups-v2 branch

[Brian Paul]

On 01/20/2012 12:45 PM, Brian Paul wrote:
> On 01/20/2012 11:56 AM, Ian Romanick wrote:
>> On 01/17/2012 02:58 PM, Brian Paul wrote:
>>> On 01/16/2012 08:30 PM, Brian Paul wrote:
>>>> On Mon, Jan 16, 2012 at 4:31 PM, Ian Romanick<idr at freedesktop.org>
>>>> wrote:
>>>>> On 01/16/2012 01:30 PM, Brian Paul wrote:
>>>>>>
>>>>>>
>>>>>> The renderbuffer-cleanups-v2 branch removes all the old swrast
>>>>>> GetRow/PutRow stuff. All swrast rendering is now done through
>>>>>> renderbuffer mapping and the format_pack/unpack.c code.
>>>>>
>>>>>
>>>>> Woo hoo! Thanks for tackling this! I've been slowly putting
>>>>> bandaids on
>>>>> some of the code, but it looks like you've saved me a ton of work!
>>>>>
>>>>>
>>>>>> The gl_renderbuffer type is smaller and cleaner now. Plus, a few
>>>>>> more
>>>>>> old driver hooks are removed.
>>>>>
>>>>>
>>>>> I'm reviewing this code, and it looks good so far. However, it looks
>>>>> like
>>>>> it won't build at every commit. That's really important for
>>>>> bisecting. For
>>>>> example, I notice the patch that adds uses of gl_renderbuffer::Map
>>>>> appear
>>>>> before the patch that adds the field.
>>>>
>>>> Hmm, I'll have to check on that tomorrow. I redid the whole patch
>>>> series today (hence -v2) and could have made a mistake along the way.
>>>
>>> I rebuilt at a few intermediate points and didn't see any issues. I
>>> tried your check_all_commits.sh script but it didn't work for me:
>>>
>>> $ ~/check_all_commits.sh . build2/
>>> f2749c52627aa1a45cbd8ee0dbbe460fbf1aeb36..496
>>> e7a2986a5abc82c982a3994b3e46d21e5a4b0
>>> /home/brian/check_all_commits.sh: 24: function: not found
>>> /home/brian/check_all_commits.sh: 37: cannot create
>>> /tmp/warn-master-.-build2/.t
>>> xt: Directory nonexistent
>>> grep: /tmp/build-master-.-build2/.txt: No such file or directory
>>> /home/brian/check_all_commits.sh: 37: cannot create
>>> /tmp/warn-f2749c52627aa1a45c
>>> bd8ee0dbbe460fbf1aeb36..496e7a2986a5abc82c982a3994b3e46d21e5a4b0-.-build2/.txt:
>>>
>>>
>>>
>>> Directory nonexistent
>>> grep:
>>> /tmp/build-f2749c52627aa1a45cbd8ee0dbbe460fbf1aeb36..496e7a2986a5abc82c982
>>>
>>>
>>> a3994b3e46d21e5a4b0-.-build2/.txt: No such file or directory
>>> Warning changes in
>>> f2749c52627aa1a45cbd8ee0dbbe460fbf1aeb36..496e7a2986a5abc82c9
>>> 82a3994b3e46d21e5a4b0 (.-bit, build2/):
>>> diff: /tmp/warn-master-.-build2/.txt: No such file or directory
>>> diff:
>>> /tmp/warn-f2749c52627aa1a45cbd8ee0dbbe460fbf1aeb36..496e7a2986a5abc82c982a
>>>
>>>
>>> 3994b3e46d21e5a4b0-.-build2/.txt: No such file or directory
>>> Installation directory does not exist.
>>>
>>> Maybe you could run the script and let me know if/where there is a bad
>>> commit.
>>
>> It looks like I was mis-reading things. I used my script to do a full
>> DRI and Gallium build with both debug and release C flags at every
>> patch, and it was fine. I also had my QA team run additional tests on
>> the branch, and there were no regressions. I seriously owe you a beer
>> for knocking out this work. Thanks.
>
> OK, I'm glad it checked out. I'll try to merge it to master later today.

I'm going to hold off on merging until Matt fixes the build problem I 
mentioned in my other message.   It'll make bisecting easier for me in 
case a regression is found.

-Brian