Fwd: [rtems-bsp-builder] 2020-10-13 09:15:22: Profile(s): everything

Sebastian Huber sebastian.huber at embedded-brains.de
Thu Oct 15 06:26:07 UTC 2020


On 15/10/2020 08:23, Chris Johns wrote:

> On 15/10/20 5:15 pm, Sebastian Huber wrote:
>> On 15/10/2020 08:05, Chris Johns wrote:
>>
>>> On 15/10/20 4:35 pm, Sebastian Huber wrote:
>>>> On 15/10/2020 00:48, Chris Johns wrote:
>>>>
>>>>> On 15/10/20 2:27 am, Joel Sherrill wrote:
>>>>>> On Wed, Oct 14, 2020 at 10:20 AM Sebastian Huber
>>>>>> <sebastian.huber at embedded-brains.de  
>>>>>> <mailto:sebastian.huber at embedded-brains.de>>
>>>>>> wrote:
>>>>>>        On 14/10/2020 17:17, Joel Sherrill wrote:
>>>>>>        > On Wed, Oct 14, 2020 at 9:45 AM Sebastian Huber
>>>>>>        > <sebastian.huber at embedded-brains.de
>>>>>>        <mailto:sebastian.huber at embedded-brains.de>
>>>>>>        > <mailto:sebastian.huber at embedded-brains.de
>>>>>>        <mailto:sebastian.huber at embedded-brains.de>>> wrote:
>>>>>>        >     On 14/10/2020 15:35, Joel Sherrill wrote:
>>>>>>        >
>>>>>>        >     > BSP builder has 81 failures. :(
>>>>>>        >     It tried to build BSPs which no longer exist.
>>>>>>        >
>>>>>>        > Well that is an easier thing to fix than my concern that it was
>>>>>> related to
>>>>>>        > giving errors when a BSP does not support a feature. I suppose that
>>>>>> will
>>>>>>        > show up when the bsp builder switches to waf.
>>>>>>        >
>>>>>>        > Can you patch ./config/rtems-bsps-tiers.ini to reflect what you
>>>>>> removed?
>>>>>>        We should try try to reduce the redundancy in our data sets. Why don't
>>>>>>        we record the tier status in the BSP items?
>>>>>>
>>>>>> That's a Chris discussion when the builder is switched to using waf.
>>>>> The current data is in rtems-tools.git/config. I needed a spot and could not
>>>>> think of another place that was easy and low impact. I would welcome
>>>>> alternatives. If this data can be generated and updated or runtime loaded from
>>>>> another source, ie spec files, I would welcome this.
>>>>>
>>>>> The current blocker with the spec files is being able to use some shared
>>>>> code to
>>>>> parse and load the YAML data plus being able to load the data quickly. Waf has
>>>>> the loading code in the wscript file so that cannot be easily shared and it
>>>>> uses
>>>>> saved pickled data which I do not think is shareable.
>>>> We have a couple of options to re-use the build specification items. In general,
>>>> the Python module to work with these items outside the wscript is in
>>>> rtems-central:
>>>>
>>>> https://git.rtems.org/rtems-central/tree/rtemsspec/items.py
>>> How long does this code take to load the build spec files in rtems.git?
>>>
>> It uses an item cache in pickle format. If you start with an empty cache it
>> needs a couple of seconds. Once the cache is set up it loads in less than half a
>> second.
> Oh nice, that functionality is part of this code? I had (incorrectly it seems)
> assumed the dependency checking was being done by waf.
The wscript uses one pickle file for the complete build specification. 
The rtemsspec/item uses one pickle file per directory. This allows 
faster updates if you just modify just one file of the specification.


More information about the devel mailing list