cpu/mips: remove periph_gpio_irq feature and stub#9974
cpu/mips: remove periph_gpio_irq feature and stub#9974cladmi merged 2 commits intoRIOT-OS:masterfrom
periph_gpio_irq feature and stub#9974Conversation
cladmi
left a comment
There was a problem hiding this comment.
ACK they were just added everywhere using sed.
cladmi
left a comment
There was a problem hiding this comment.
Ooops, just checked, the hifive1 board uses fe310 cpu that defines gpio_irq_enable.
805e749 to
1ec3a2b
Compare
Even if so, I would expect at least the reviewers to notice these kind of things! Its just very annoying to have these simple fails all over the place all the time... |
|
removed the patch for the |
|
The |
Sorry, but that is bullshit (and very unclean). |
|
Murdock checks that |
|
To keep a working git history, the commits should be in the inverted order with disabling the feature before removing the code. |
|
Yes, but this is not about making Murdock pass somehow, but about using the feature defs as they are meant to: declare, if a certain feature is available on a certain platform! I don't mean to be rude, its just that I think we need to be more careful when reviewing/refactoring/applying API changes, as we tend to be way to sloppy when doing these things... So next time lets look at these kind of changes more carefully! |
Last time I checked the convention was, that not every commit needs to be able to build... But I can adapt this. |
1ec3a2b to
da899da
Compare
|
changed order of commits. |
cladmi
left a comment
There was a problem hiding this comment.
ACK waiting for murdock.
Github is still displaying them in the old order but it is a known issue with github displaying by 'author' date, checking locally displays the correct commit order.
|
@haukepetersen it helps if every commit can build properly when doing bisections. It may not always be desirable due to the resulting size of a single commit in certain cases, but try to keep the build working for each commit if possible. |
|
yepp, makes sense. |
Contribution description
The
gpiodriver for thepic32does not implement external interrupts (feature ´periph_gpio_irq`), it had simply a stub function for it. With #9035 the clean way to handle this is however to remove the stub and the feature declaration from the boards using that CPU, so applications that depend no this GPIO feature do not event build for those boards anymore...Testing procedure
build #9035 with and without this PR
Issues/PRs references
Stumbled upon this while finalizing #9035
#9845 broke it :-)