ioreg syntax extension #108

bgamari · 2014-07-17T05:11:20Z

I'll need to write something more coherent here but here is the continuation of the ioreg work originally discussed in #90. I've restructured and cleaned up the builder code substantially and getter and setter objects have been implemented. Note that this currently requires some compiler bits that have yet to be merged to compile. You can find rustdoc output for the code generated by a simple example here. More on this later when it's not as late.

I would apply the proper label to indicate this is work in progress but it seems I don't have permission to do so.

Remaining work

Examine options for reducing documentation clutter

bgamari · 2014-07-17T14:46:41Z

@bharrisau I've updated the rustdoc upload, feel free to have a look. The FTM_STATUS register has set_to_clear flags.

farcaller · 2014-07-17T16:50:50Z

ioreg/builder/accessors.rs

+use super::super::node;
+
+/// A visitor to build accessor functions for each register struct
+pub struct BuildAccessors<'a, 'b, 'c> {


Any reason not to unify lifetimes here?

Vladimir Pouzanov [email protected] writes:

+/// A visitor to build accessor functions for each register struct
+pub struct BuildAccessors<'a, 'b, 'c> {

Any reason not to unify lifetimes here?

I played around with these for quite a while to make the borrow checker
happy. That being said, things have changed substantially since then, so
perhaps they can now be unified. I'll have a look later.

farcaller · 2014-07-17T16:58:02Z

I've added a few notes on the code. I will have some more on the style later on when this settles a bit, but overall it's very nice, great job!

bgamari · 2014-07-30T00:24:14Z

@farcaller, @bharrisau what do we think about peripheral instantiation? I'm tempted to punt on bit-banding support for now as I don't have a lot of time recently and I'd like to get this merged eventually.

bharrisau · 2014-07-30T00:27:43Z

I'll throw bit-banding in later. The easiest is to use weak linkage of the bitband memory address (and I'll be able to do a memory map macro when I get a chance that will generate that).

What do you mean by peripheral instantiation? As in, generating the #[link_name="armmem_SYSTICK"] pub static SYSTICK: SYSTICKReg line? I'd leave that out, and we can do it in platformtree or a memory map macro. (I'd like to shift the device specific stuff one level higher than HAL - memory map and pin map. Leave HAL to handle the vendor/architecture specific stuff)

bgamari · 2014-07-30T02:19:05Z

@bharrisau That sounds reasonable enough.

Unfortunately it seems like something has broken with the recent rustc/zinc changes. After rebasing I'm seeing the following unhelpful build failure,

rustc --opt-level 2 -Z no-landing-pads --cfg cfg_mcu_has_spi --cfg mcu_lpc17xx --cfg arch_cortex_m3 --target thumbv7m-linux-eabi -Ctarget-cpu=cortex-m3 -C relocation_model=static      -L /opt/exp/zinc/build  --out-dir /opt/exp/zinc/build  /opt/exp/zinc/src/main.rs 
/opt/exp/zinc/src/main.rs:1:1: 1:1 error: expected ident, found `=`
/opt/exp/zinc/src/main.rs:1 // Zinc, the bare metal stack for rust.

bharrisau · 2014-07-30T02:57:04Z

If I had to guess - there is an '=' used incorrectly in a macro.

bharrisau · 2014-07-30T02:58:07Z

Maybe somewhere you are calling .expect_ident?

farcaller · 2014-07-30T07:16:48Z

No, this is rust-lang/rust#15972. Build with -O0.

farcaller · 2014-07-30T07:17:06Z

(I've seen the same problem in #116)

bgamari · 2014-07-30T14:38:40Z

@farcaller it seems like the fix has made it upstream. I'm building a new rustc as we speak. Well done tracking this one down, by the way.

bgamari · 2014-08-06T17:43:28Z

This morning I brushed off these patches, rebased them, and did a bit more porting of K20. I found a few bugs and cleaned up the documentation a bit further in the process. At this point I feel pretty good about merging so folks can finally start using this in anger.

What do people think?

farcaller · 2014-08-06T18:04:29Z

ioreg/builder/accessors.rs

+// limitations under the License.
+
+use syntax::ast;
+use syntax::ast::{P};


Remove {/} here and on next line.

farcaller · 2014-08-06T18:30:53Z

Thanks for all the work, looks nice.

I've left a few comments in for you. Also, can you split the macro part and the new ioreg implementations to separate PRs?

bgamari · 2014-08-06T19:35:33Z

I almost forgot: My tree currently has deny(missing_doc) commented out as many of the register definitions don't have documentation. In general, I think we should accept a slightly lower standard of documentation for the register interfaces than the high-level library interfaces. The register interfaces have fewer users and frankly I think it's reasonable to expect that you'll need to datasheet by your side when you are poking at registers.

Would there be any objection to my adding allow(missing_doc) annotations to items produced by ioregs!?

farcaller · 2014-08-06T20:08:31Z

Would there be any objection to my adding allow(missing_doc) annotations to items produced by ioregs!?

I'm fine with that. It might be useful to add a reference to where in the datasheet you can find the register description, but those fluctuate quite as well. Maybe chapter number / name would be fine?

bgamari · 2014-08-06T22:13:58Z

@farcaller as far as I can tell things look good for merge.

errordeveloper · 2014-08-06T22:23:06Z

@bgamari great stuff! I don't seem to see any unit tests here, so I suppose you are intending to add them in a separate pull request...

bgamari · 2014-08-06T22:29:51Z

@errordeveloper there is still a bit of work to be done on the documentation and testing side of things but I'd like to get this merged to get people using it and lessen the maintenance burden.

Need to define library path; not sure how this worked before.

bharrisau · 2014-08-07T04:05:00Z

ioreg/builder/setter.rs

+        let clear_mask: $unpacked_ty = $clear as $unpacked_ty;
+        if self.mask != 0 {
+          let v: $unpacked_ty = self.reg.value.get() & ! clear_mask & ! self.mask;
+          self.reg.value.set(self.value | v);


As some future work, will just double check how the generated ASM for this looks.

bharrisau · 2014-08-07T04:17:38Z

Wow, awesome work. I haven't had a chance to have a thorough look (and I probably wont until my contract finishes at the end of this month).

Can someone land a PR that renames all the current ioreg! to ioreg_old! to make room for this one?

farcaller · 2014-08-07T07:05:49Z

Ok, let's land this.

ioreg syntax extension Reviewed-by: farcaller

Added new ioreg syntax extension

bharrisau added the in progress label Jul 17, 2014

farcaller reviewed Jul 17, 2014
View reviewed changes

errordeveloper mentioned this pull request Jul 27, 2014

Analog components #117

Open

farcaller reviewed Aug 6, 2014
View reviewed changes

bgamari added 6 commits August 6, 2014 16:56

Initial commit of ioreg! syntax extension

0d9a585

ioreg: Mark accessors with allow(missing_doc)

8ede9f3

ioreg::setter: Fix clear flag docstring

83342c9

ioreg::accessors: Add docstring to get()

dce7673

ioreg: Allow enums to lack documentation

3c80963

ioreg: Allow register unions to lack documentation

6729d29

bgamari added 3 commits August 6, 2014 20:22

Rakefile: Fix build_docs rule

2c7e926

Need to define library path; not sure how this worked before.

Rakefile: Add ioreg to build_docs

f3c4f52

ioreg: Add documentation

bf1fe45

bharrisau reviewed Aug 7, 2014
View reviewed changes

hacknbot added a commit that referenced this pull request Aug 7, 2014

Merge pull request #108 from bgamari/ioreg-rebased

6f6bb5b

ioreg syntax extension Reviewed-by: farcaller

hacknbot added a commit that referenced this pull request Aug 7, 2014

Merge pull request #108 from bgamari/ioreg-rebased

c046956

ioreg syntax extension Reviewed-by: farcaller

hacknbot added a commit that referenced this pull request Aug 7, 2014

Merge pull request #108 from bgamari/ioreg-rebased

5955517

ioreg syntax extension Reviewed-by: farcaller

hacknbot added a commit that referenced this pull request Aug 7, 2014

Merge pull request #108 from bgamari/ioreg-rebased

7bcf91c

ioreg syntax extension Reviewed-by: farcaller

farcaller added a commit that referenced this pull request Aug 7, 2014

Merge pull request #108 from bgamari/ioreg-rebased

51e2abc

Added new ioreg syntax extension

farcaller merged commit 51e2abc into hackndev:master Aug 7, 2014

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ioreg syntax extension #108

ioreg syntax extension #108

bgamari commented Jul 17, 2014

bgamari commented Jul 17, 2014

farcaller Jul 17, 2014

bgamari Jul 17, 2014

farcaller commented Jul 17, 2014

bgamari commented Jul 30, 2014

bharrisau commented Jul 30, 2014

bgamari commented Jul 30, 2014

bharrisau commented Jul 30, 2014

bharrisau commented Jul 30, 2014

farcaller commented Jul 30, 2014

farcaller commented Jul 30, 2014

bgamari commented Jul 30, 2014

bgamari commented Aug 6, 2014

farcaller Aug 6, 2014

farcaller commented Aug 6, 2014

bgamari commented Aug 6, 2014

farcaller commented Aug 6, 2014

bgamari commented Aug 6, 2014

errordeveloper commented Aug 6, 2014

bgamari commented Aug 6, 2014

bharrisau Aug 7, 2014

bharrisau commented Aug 7, 2014

farcaller commented Aug 7, 2014

ioreg syntax extension #108

ioreg syntax extension #108

Conversation

bgamari commented Jul 17, 2014

Remaining work

bgamari commented Jul 17, 2014

farcaller Jul 17, 2014

Choose a reason for hiding this comment

bgamari Jul 17, 2014

Choose a reason for hiding this comment

farcaller commented Jul 17, 2014

bgamari commented Jul 30, 2014

bharrisau commented Jul 30, 2014

bgamari commented Jul 30, 2014

bharrisau commented Jul 30, 2014

bharrisau commented Jul 30, 2014

farcaller commented Jul 30, 2014

farcaller commented Jul 30, 2014

bgamari commented Jul 30, 2014

bgamari commented Aug 6, 2014

farcaller Aug 6, 2014

Choose a reason for hiding this comment

farcaller commented Aug 6, 2014

bgamari commented Aug 6, 2014

farcaller commented Aug 6, 2014

bgamari commented Aug 6, 2014

errordeveloper commented Aug 6, 2014

bgamari commented Aug 6, 2014

bharrisau Aug 7, 2014

Choose a reason for hiding this comment

bharrisau commented Aug 7, 2014

farcaller commented Aug 7, 2014