[Mesa-dev] [PATCH v3 00/15] cleanup anv_entrpoints_gen.py

[Emil Velikov]

On 27 February 2017 at 19:36, Emil Velikov <emil.l.velikov at gmail.com> wrote:
> On 27 February 2017 at 19:20, Dylan Baker <dylan at pnwbakers.com> wrote:
>> Quoting Emil Velikov (2017-02-26 10:26:24)
>>> On 24 February 2017 at 18:21, Dylan Baker <dylan at pnwbakers.com> wrote:
>>> > There are a number of small style cleanups and simplifications in this series,
>>> > but the main changes are:
>>> >  - use a mako template to generate the header and code rather than prints
>>> >  - be python 3.x ready (the goal isn't to write python 3 code, but to write code
>>> >    that is easy to port or hybridize)
>>> >  - generate the header and the code in one go
>>> >
>>> > I've put emphasis on the readability of the template rather than the readability
>>> > of the output code, it's relatively easy to pipe the code through 'indent' to
>>> > make it more readable.
>>> >
>>> > Notable changes in Version 2:
>>> > - Pass XML file via an argument
>>> > - add flag to control output directory
>>> > - Attempt to update android makefiles
>>> >
>>> > Notable changes in Version 3:
>>> > - Fix "do not edit" message to have proper file name
>>> > - More Android.mk changes
>>>
>>> > - Don't write both files at the same time
>>> > - Provide the file name to be written to as an argument
>>> >
>>> Not sure why you opted to rewrite this considering that we already
>>> have such bugs in-tree and the fix was a 2-line change in the
>>> makefile(s). Perhaps I should have made that clearer :-(
>>
>> Eric seemed to suggest that this approach was impossible in android makefiles. I
>> can revert that change easily if that's not the case.
>>
>>>
>>> That aside - tried to pull/test your series, but patchwork has gone
>>> crazy (doesn't like git series?). Do you have a branch somewhere ?
>>
>> I can push a branch, just let me know whether I'm wrong about the android
>> makefile issue first.
>>
> The issue in question is more often seen in the Android makefiles.
> I've poked the guys a few times to actually use the generic make ones
> that we use in Autoconf.
> If they're don't have time/interest [it is hard to trigger after all]
> I won't bother you with any of it.
>
Completely forgot - please ensure that the script fails early if the
arguments are not provided. I.e. we need some "Required = True" in the
parser.

Thanks
Emil