Since _imp isn't directly exposed it isn't a priority to do the conversion, but for maintenance and such it should still occur.

I'm normalizing the names of these issues so they're easier to read and to search for.

I have attached my conversion of import.c. Larry, can you look it over to make sure I did it as expected? I tried to not have to write my own converter for O& as the docs seem to suggest object(converter='...') should work, but I got an error instead.

(Brett: in order to avoid creating 129 differet issues on the tracker, I'm bundling files together into groups of 50 sites.  You don't have to tackle the whole list if you don't want to.)

This issue is part of the Great Argument Clinic Conversion Derby,
where we're trying to convert as much of Python 3.4 to use
Argument Clinic as we can before Release Candidate 1 on January 19.

This issue asks you to change the following bundle of files:
    Python/import.c: 8 sites
    Modules/cmathmodule.c: 8 sites
    Modules/cjkcodecs/multibytecodec.c: 8 sites
    Modules/arraymodule.c: 8 sites
    Modules/pyexpat.c: 7 sites
    Modules/fcntlmodule.c: 7 sites
    Modules/pwdmodule.c: 2 sites
    Modules/spwdmodule.c: 1 sites
    Modules/grpmodule.c: 1 sites

Talk to me (larry) if you only want to attack part of a bundle.

For instructions on how to convert a function to work with Argument
Clinic, read the "howto":
    http://docs.python.org/dev/howto/clinic.html

New changeset 1f3242fb0c9c by Brett Cannon in branch 'default':
Issue #20152: import.c now uses Argument Clinic.
http://hg.python.org/cpython/rev/1f3242fb0c9c

Now it's:

    Modules/cmathmodule.c: 8 sites
    Modules/cjkcodecs/multibytecodec.c: 8 sites
    Modules/arraymodule.c: 8 sites
    Modules/pyexpat.c: 7 sites
    Modules/fcntlmodule.c: 7 sites
    Modules/pwdmodule.c: 2 sites
    Modules/spwdmodule.c: 1 sites
    Modules/grpmodule.c: 1 sites

I'm going to start tackling some of these today and checking them in directly. If I don't finish I will update the bug with what I didn't get to.

Patch for multibytecodec.c.

Not sure how you calculated locations, Larry, but this file had 11 things to change, so some other work divisions might be off.

After finding out that the initial site counts were off I went through and manually counted each file in this Derby group to get a more accurate idea of how much work each file would be.

    Modules/cmathmodule.c: 8 sites
    Modules/arraymodule.c: 27 sites
    Modules/pyexpat.c: 11 sites
    Modules/fcntlmodule.c: 4 sites
    Modules/pwdmodule.c: 3 sites
    Modules/spwdmodule.c: 2 sites
    Modules/grpmodule.c: 3 sites

Might pick this up later today (or never), but taking a break for now.

Larry, I have patches for grp, spwd, _multibytecodec, and pwd I'd like you to give a once-over to make sure I didn't miss anything.

The only one that I think has an obvious improvement is in pwd where there is custom logic to deal with a PyArg_ParseTuple() failure. Does Argument Clinic have a way to specify custom logic in that instance?

With these patches the TODO for this derby group is:

    Modules/cmathmodule.c: 8 sites
    Modules/arraymodule.c: 27 sites
    Modules/pyexpat.c: 11 sites
    Modules/fcntlmodule.c: 4 sites

cmath might not be worth doing, though, for 3.4 as it already has custom macros to create function definitions.

Here is fcntl converted. I didn't do fcntl.ioctl as it has a crazy set of possible signatures which I simply didn't want to deal with.

Here is pyexpat converted. Couldn't do ErrorString as the 'l' format isn't supported; issue #20332.

Here is pyexpat updated with ErrorString converted (and generally updated to manage changes now required by AC for class definitions).

Should only have two modules left to convert:

    Modules/cmathmodule.c: 8 sites
    Modules/arraymodule.c: 27 sites

I have already started arraymodule.c; it's just taking a while.

It took a bit of finessing but I managed to convert cmath in a way that didn't make it worse compared to before AC.

I now consider this part of the derby done and ready for Larry to review.

multibytecodec_derby.diff looks fine, please apply.

The cmath patch fails to apply; please update it.

New changeset c2d71f9afa0b by Brett Cannon in branch 'default':
Issue #20152: Convert _multibytecodecs to Argument Clinic.
http://hg.python.org/cpython/rev/c2d71f9afa0b

To make it easier to track what is left to commit:

* grp
* spwd
* fcntl
* pyexpat
* array
* cmath [Martin says this no longer applies cleanly]

I should also mention all of these patches probably need to shift to `output preset file` as well.

New changeset 2617db7e43a3 by Brett Cannon in branch 'default':
Issue #20152: Convert the grp module to Argument Clinic.
http://hg.python.org/cpython/rev/2617db7e43a3

To make it easier to track what is left to commit:
* spwd
* fcntl
* pyexpat
* array
* cmath [Martin says this no longer applies cleanly]

New changeset b3aa30f474c4 by Brett Cannon in branch 'default':
Issue #20152: Port the spwd module to Argument Clinic.
http://hg.python.org/cpython/rev/b3aa30f474c4

New changeset c48980af7df2 by Brett Cannon in branch 'default':
Issue #20152: Port the pwd module to Argument Clinic.
http://hg.python.org/cpython/rev/c48980af7df2

New changeset 0c2792a8f9db by Brett Cannon in branch 'default':
Issue #20152: Port pyexpat to Argument Clinic.
http://hg.python.org/cpython/rev/0c2792a8f9db

To make it easier to track what is left to commit:
* fcntl [does not apply cleanly
* array
* cmath [Martin says this no longer applies cleanly]

New changeset f7ab0884467e by Brett Cannon in branch 'default':
Issue #20152: Port the array module to Argument Clinic.
https://hg.python.org/cpython/rev/f7ab0884467e

Only ones left are:

- fcntl
- cmath

Both no longer apply cleanly.

New changeset 5e8b94397f81 by Brett Cannon in branch 'default':
Issue #20152: Convert the cmath module to Argument Clinic.
https://hg.python.org/cpython/rev/5e8b94397f81

Here is a new patch for fcntl. I would like a review since I had to get a bit tricky to handle the polymorphic arguments for fcntl and ioctl and I don't think the test suite is that thorough for this module.

I found bugs or doubtful code in the fcntl module. They should be fixed before converting to Argument Clinic.

So I disagree that the code needs to be tweaked before converting to Argument Clinic. If the Clinic conversion is not adding to the problem then the code churn is just going to make applying this patch that much harder.

Thanks for the code review regardless, though!

If first convert to Argument Clinic then fixing bugs will be much harder.

About argument names. You have changed argument names and docstrings in any case (e.g. was "op", now "code"). Why not conform with standard documentation? This wouldn't add additional code churn if change it now. But will add if change it later.

New changeset 6e6532d313a1 by Brett Cannon in branch 'default':
Issue 20152, 22821: Port the fcntl module to Argument Clinic.
https://hg.python.org/cpython/rev/6e6532d313a1

Finally finished!

I am never trusting Larry to count anything ever again. ;)