[PATCH] [CAGE] Perl-to-PIR conversion of two tests in the t/library/ directory

[geraud]

Patch attached. I'm not really sure about the unicode section in data_escape.t, so comments are welcomed.

[jkeenan]

GeJ,

Can you clarify the rationale for these conversion? (I may have missed some discussion of this somewhere.)

Thank you very much.
kid51

[coke]

One of the testing "nice to haves" (I hesitate to call it a goal) is to translate the test that are amenable to it from perl-based to PIR based.

The two main benefits here are "eat our own dogfood", and improved speed on the tests.

See https://trac.parrot.org/parrot/wiki/ConvertTestsToParrot for the How (but not the Why)

[geraud]

James notified me on IRC that the patch didn't apply correctly for him. I tried to apply it locally on a fresh copy and it indeed failed. Looking at the rejection file, it seems to be $Id$ related. Am I missing something else?

[jkeenan]

Replying to geraud:

Am I missing something else?

I think the problem is in patch, not in you or me. For the purpose of moving things forward, I applied your patch, then edited the two resulting files so that the .rej matter was correctly positioned. I then tested them, first individually with prove and then with make test. See  Smolder test report here.

Then, I did a bit more research on the patch problem. I RTFM for patch, which had nothing specific to this problem. From the work above, I created an svn diff, did a fresh svn co into another working copy, applied my patch with patch just as I had with yours -- and got exactly the same .rej results. So this is a limitation in patch. I speculate that it may have to do with the fact that we're changing the first line of the file -- but that's merely speculation.

So the tests are now in PIR and they're all passing. But the files should still be reviewed by someone with more PIR-fu than me. And I can't shed much light on your Unicode questions.

Thank you very much.
kid51

[geraud]

Replying to jkeenan:

I think the problem is in patch, not in you or me.

I would have turned the problem differently: I think the issue lies with how svn diff works (in our case not expanding svn-specific keywords properly) which makes the resulting output not patch(1)-friendly.

This is a known  annoyance in Subversion for which the only known workaround seems to be :

"adjusting the diff by hand."

I'll try to be more careful next time and generate a diff that will apply correctly. Sorry for the trouble I caused and thanks for the commit.

Géraud

[jkeenan]

Can we get a code review here? Particular concerns: Was the Perl 5-to-PIR conversion done correctly? Any problems with the Unicode?

Thanks.
kid51

[geraud]

Here is a short summary (a before-after version) of the Unicode tests to ease the job of the reviewer :  http://nopaste.snit.ch/15291

My concerns are mostly on how I generate the unicode strings. My solution is the _unicode_gen sub that takes one string parameter (one of "0666", "0777", etc.) and returns the unicode:"\u0666" (unicode:"\u0777", ...) via an eval-like ( compreg "PIR" then compile the code and call it ). If there's an easier way to achieve this goal, do not hesitate to fix the test.

Thanks in advance, Géraud

[jkeenan]

No one has complained. The tests keep passing. We'll close this ticket for now.

Thank you very much.
kid51