makefiles/info-global: Remove dependency resolution cache

[leandrolanzieri]

Contribution description

This removes a preliminar dependency resolution that is performed without including features, to skip a full dependency resolution when boards can be proven to be unsupported on an earlier stage.

This was introducing issues on some boards since the blacklisting of some features depends on board information which is not available at the time of performing this resolution.

Some time measurements

Running make info-boards-supported 50 times on my computer results in:

Application	Time with cache	Time without cache
hello_world	873.0 ms ± 20.5 ms	851.7 ms ± 26.3 ms
tests/driver_ws281x (proposed here)	266.4 ms ± 11.6 ms	263.0 ms ± 18.6 ms

Testing procedure

• make info-boards-supported should work properly now (it was pointed out in the mailing list that the sodaq boards are not showing up in the list in current master, for instance)

Issues/PRs references

Spotted by @fjmolinas here

[maribu]

Is it correct, that it is faster without caching? Or is this columns mixed up?

In any case, having this working is a higher priority than having it fast. So even if it would slow things down, we obviously have to accept this.

[leandrolanzieri]

Is it correct, that it is faster without caching? Or is this columns mixed up?

No, the columns are Ok. When running the benchmark, without the cache took less time.

[maribu]

OK, with this the following additional boards are listed in examples/hello-world:

• adafruit-cluearduino-mkr1000
• arduino-mkrfox1200
• arduino-mkrwan1300
• arduino-mkrzero
• feather-m0
• nrf52840dongle
• serpente
• sodaq-autonomo
• sodaq-explorer
• sodaq-one
• sodaq-sara-aff

I compiled it for each of these, it worked for all.

[maribu]

ACK

[leandrolanzieri]

Thanks for the quick review.

[maribu]

When running the benchmark, without the cache took less time.

Then it is extra good to have this in :-)

(I'm sure it was noticeably faster with the cache on my machine at the time I added this. I wonder why it turns out slowing down things now.)

[fjmolinas]

@keestux the issue you mentioned on the mailing list should be fixed now with this PR.

[kaspar030]

does this need a backport? @leandrolanzieri

[leandrolanzieri]

does this need a backport? @leandrolanzieri

Hmm yes, you're right

[leandrolanzieri]

Backport provided in #14136