Version dependent error messages #643

abonie · 2017-08-20T09:42:50Z

There are multiple discrepancies in error messages for Python 3.5+, which causes 3.5 and 3.6 test suites to fail. This tries to fix them.

abonie · 2017-08-20T10:06:13Z

Anyone knows how to handle situation where for certain test cases there are expected failures in Python 3.4 but unexpected successes for Python 3.5?
@freakboy3742

hgluka · 2017-08-20T13:54:57Z

How many of those instances are there? I only see 20 failures in the output from BeeKeeper (no unexpected successes).
Anyway, you might want to make a special decorator that handles those cases. It's only a matter of adding a version check to expectedFailure.

brutusthebee · 2017-08-20T14:19:56Z

batavia/VirtualMachine.js

+        var args
+        var extra = 0
+        var lo
+        var hi


At column 13: (no-redeclare) 'pos' is already defined.

brutusthebee · 2017-08-20T14:19:56Z

batavia/VirtualMachine.js

+        var extra = 0
+        var lo
+        var hi
+


At column 13: (no-redeclare) 'unpacked_code' is already defined.

brutusthebee · 2017-08-20T14:19:57Z

batavia/VirtualMachine.js

+        var lo
+        var hi
+
+        while (pos < code.co_code.val.length) {


At column 13: (no-redeclare) 'args' is already defined.

brutusthebee · 2017-08-20T14:19:57Z

batavia/VirtualMachine.js

+        var hi
+
+        while (pos < code.co_code.val.length) {
+            var opcode_start_pos = pos


At column 13: (no-redeclare) 'extra' is already defined.

brutusthebee · 2017-08-20T14:19:58Z

batavia/VirtualMachine.js

+            var opcode = code.co_code.val[pos++]
+
+            // next opcode has 4-byte argument effectively.
+            if (opcode === dis.EXTENDED_ARG) {


At column 17: (no-redeclare) 'opcode_start_pos' is already defined.

brutusthebee · 2017-08-20T14:19:58Z

batavia/VirtualMachine.js

+            // next opcode has 4-byte argument effectively.
+            if (opcode === dis.EXTENDED_ARG) {
+                lo = code.co_code.val[pos++]
+                hi = code.co_code.val[pos++]


At column 17: (no-redeclare) 'opcode' is already defined.

brutusthebee · 2017-08-20T14:19:58Z

batavia/VirtualMachine.js

+
+                if (opcode in dis.hasconst) {
+                    args = [code.co_consts[intArg]]
+                } else if (opcode in dis.hasfree) {


At column 21: (no-redeclare) 'intArg' is already defined.

brutusthebee · 2017-08-20T14:19:59Z

batavia/VirtualMachine.js

+                    }
+                } else if (opcode in dis.hasname) {
+                    args = [code.co_names[intArg]]
+                } else if (opcode in dis.hasjrel) {


At column 29: (no-redeclare) 'var_idx' is already defined.

brutusthebee · 2017-08-20T14:19:59Z

batavia/VirtualMachine.js

+    this.push(dict)
+    return
+}
+
 VirtualMachine.prototype.byte_STORE_MAP = function() {
    switch (constants.BATAVIA_MAGIC) {
        case constants.BATAVIA_MAGIC_35:


At column 5: (no-useless-return) Unnecessary return statement.

brutusthebee · 2017-08-20T14:20:00Z

batavia/VirtualMachine.js

+        var lenPos = arg % 256
+        for (var i = 0; i < lenKw; i++) {
+            var items = this.popn(2)
+            namedargs[items[0]] = items[1]


At column 13: (no-redeclare) 'namedargs' is already defined.

brutusthebee · 2017-08-20T14:20:00Z

batavia/VirtualMachine.js

+            var items = this.popn(2)
+            namedargs[items[0]] = items[1]
+        }
+        if (kwargs) {


At column 13: (no-redeclare) 'lenPos' is already defined.

brutusthebee · 2017-08-20T14:20:01Z

batavia/VirtualMachine.js

+            for (let elem of args) {
+                posargs.push(elem)
+            }
+        }


At column 13: (no-redeclare) 'posargs' is already defined.

brutusthebee · 2017-08-20T14:20:01Z

batavia/VirtualMachine.js


-    var retval = func(posargs, namedargs)


At column 13: (no-redeclare) 'func' is already defined.

brutusthebee · 2017-08-20T14:20:02Z

batavia/VirtualMachine.js

-    this.push(retval)
+        var retval = func(posargs, namedargs)
+        this.push(retval)
+    }
 }

 VirtualMachine.prototype.byte_RETURN_VALUE = function() {


At column 13: (no-redeclare) 'retval' is already defined.

abonie · 2017-08-20T22:18:23Z

Thanks lucantrop. There were some unexpected successes, they just did not appear at the bottom.

And it turns out there already was a solution for this problem in not_implemented_versions :)

phildini · 2017-08-23T05:51:47Z

This is pretty incredible, and you've got a 🚢 from me.

It's a shame we have to repeat the <3.6 cases so many times. If you felt like condensing that logic at all, that would be fascinating and most likely very helpful.

abonie · 2017-08-23T08:00:08Z

@phildini You mean that CI task for Python 3.4 has to be run each time, even though PR regards Python 3.6?

Well, I am not well-versed with Ops part of DevOps, so one thing that comes to my mind would be parsing commit/PR message by BeeKeeper, looking for some special line like "# test Python 3.6" and changing tasks based on that. This sounds like a mess though, and also note that:

Ultimately we would like to have separate branches for different Python versions, and then this problem is easily solved via beekeeper.yml config file.
As long as Batavia supports different versions on a single branch, there is a risk, that accidentally someone will mess up how it behaves for Python 3.4 while working on Python 3.5, so running all tests may actually be necessary :(

Python 3.6+

There is an option to specify in which Python versions certain test cases would fail. Before this commit, versions would be specified as floats. This changes it from floats to strings, which seems more intuitive.

phildini · 2017-08-23T16:35:04Z

Sorry, I should have been clearer!

What I meant was we do

case constants.BATAVIA_MAGIC_34:
case constants.BATAVIA_MAGIC_35a0:
case constants.BATAVIA_MAGIC_35:
case constants.BATAVIA_MAGIC_353:
...
case constants.BATAVIA_MAGIC_36:
...

a whole lot in these changes, which seems like duplicating a lot of code? Also maybe brittle for supporting future versions.

If there were some helper functions to deal with this, it would make me happy, but this whole complaint is more "nit" than "blocker", so feel free to say you don't want to action that right now and we can probably still move forward.

abonie · 2017-08-23T16:50:28Z

@phildini Oh, sure. I have actually already started working on a module that would allow something more like this:

if (version.older('3.6')) {
    ...
} else {
    ...
}

I thought I'd submit it in separate patch though, I am not sure if I can finish it before Monday, since Thursday through Sunday I won't be able to do any work (and it's not super straight forward as there are some caveats with pre/postreleases).

Also this too goes away once we move to one branch per version.

phildini · 2017-08-23T17:06:43Z

Fair enough! If you get to the point where tests are passing on this branch, I'm happy to merge in and save the optimization for post-merge.

abonie · 2017-08-23T17:39:34Z

Great! BTW I don't know what's going on with CI right now, seems like it crashed on Python lint check..?

phildini · 2017-08-23T20:46:37Z

I'm not entirely sure what's going on either... could you try rebasing against master and pushing again?

phildini · 2017-08-26T22:35:23Z

@abonie do you feel this is ready to merge? I'm happy to do so if you're ready.

phildini · 2017-08-26T22:36:08Z

batavia/builtins/bytes.js

@@ -26,7 +27,19 @@ function bytes(args, kwargs) {
    } else if (args.length === 1) {
        var arg = args[0]
        if (arg === null) {
-            throw new exceptions.TypeError.$pyclass("'NoneType' object is not iterable")
+            switch (constants.BATAVIA_MAGIC) {


It would be great if we could condense these cases down to a helper function, but I'm fine with that being a cleanup PR later.

math module's docstring changed in Python 3.5. For now it's ok to mark respective tests as failing.

abonie · 2017-08-27T21:11:34Z

@phildini I made one last change or rather retracted one of the earlier commits, as it seems that error message in sorted has been changed in Python 3.6.2 (which I had on my machine), but BeeKeeper runs Python 3.6, so that change actually caused a failure. It is definitely ready to be merged. Failing test case in test_str.py in FormatTests seems like more than just an error message discrepancy.

brutusthebee reviewed Aug 20, 2017

View reviewed changes

batavia/VirtualMachine.js Outdated

var args

var extra = 0

var lo

var hi

Copy link

brutusthebee Aug 20, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At column 13: (no-redeclare) 'pos' is already defined.

brutusthebee reviewed Aug 20, 2017

View reviewed changes

batavia/VirtualMachine.js Outdated

var extra = 0

var lo

var hi

Copy link

brutusthebee Aug 20, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At column 13: (no-redeclare) 'unpacked_code' is already defined.

brutusthebee reviewed Aug 20, 2017

View reviewed changes

batavia/VirtualMachine.js Outdated

var retval = func(posargs, namedargs)

Copy link

brutusthebee Aug 20, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At column 13: (no-redeclare) 'func' is already defined.

brutusthebee reviewed Aug 20, 2017

View reviewed changes

abonie force-pushed the version_dependent_error_messages branch 5 times, most recently from 9192c4d to a6caeb3 Compare August 20, 2017 20:35

abonie force-pushed the version_dependent_error_messages branch 4 times, most recently from 8675062 to 808bffc Compare August 22, 2017 17:49

abonie changed the title ~~[WIP] Version dependent error messages~~ Version dependent error messages Aug 22, 2017

abonie mentioned this pull request Aug 22, 2017

GSoC: JSON module, generators and context managers #483

Open

abonie added 11 commits August 23, 2017 11:08

Fix error message in complex() for Python 3.5.3+

4bdd32e

Fix error for chr() and Python 3.6+

411df8c

Fix error message in bytes for Python 3.6+

f4315ab

Fix unorderable types error messages

5ce9f26

Fix error message for bytes __add__ method

9f8e764

Python 3.6+

Linting issues

dd898e0

Change how not_implemented_versions is used

4e41bca

There is an option to specify in which Python versions certain test cases would fail. Before this commit, versions would be specified as floats. This changes it from floats to strings, which seems more intuitive.

Fix error message for string __add__ Python 3.6+

9e4dd74

Fix few version differences in json module

5db7cdc

Small changes in few tests

a9a3692

Change format error message for Python 3.6

4da26cf

abonie force-pushed the version_dependent_error_messages branch 2 times, most recently from e7cd63c to c07cde4 Compare August 23, 2017 13:17

abonie mentioned this pull request Aug 23, 2017

Google Summer of Code 2017 final submission beeware/beeware.github.io#183

Merged

abonie force-pushed the version_dependent_error_messages branch from c07cde4 to 603c4c9 Compare August 23, 2017 20:58

phildini approved these changes Aug 26, 2017

View reviewed changes

abonie force-pushed the version_dependent_error_messages branch from 603c4c9 to 82a515f Compare August 27, 2017 18:53

Mark docstring tests as expected failure

2a6cec1

math module's docstring changed in Python 3.5. For now it's ok to mark respective tests as failing.

abonie force-pushed the version_dependent_error_messages branch from 82a515f to 2a6cec1 Compare August 27, 2017 19:18

phildini merged commit c85d032 into beeware:master Aug 30, 2017

abonie mentioned this pull request Sep 6, 2017

[WIP] Runtime version checking #654

Merged

4 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Version dependent error messages #643

Version dependent error messages #643

abonie commented Aug 20, 2017

abonie commented Aug 20, 2017 •

edited

Loading

hgluka commented Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

brutusthebee Aug 20, 2017

abonie commented Aug 20, 2017

phildini commented Aug 23, 2017

abonie commented Aug 23, 2017

phildini commented Aug 23, 2017

abonie commented Aug 23, 2017 •

edited

Loading

phildini commented Aug 23, 2017

abonie commented Aug 23, 2017

phildini commented Aug 23, 2017

phildini commented Aug 26, 2017

phildini Aug 26, 2017

abonie commented Aug 27, 2017

Version dependent error messages #643

Version dependent error messages #643

Conversation

abonie commented Aug 20, 2017

abonie commented Aug 20, 2017 • edited Loading

hgluka commented Aug 20, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

abonie commented Aug 20, 2017

phildini commented Aug 23, 2017

abonie commented Aug 23, 2017

phildini commented Aug 23, 2017

abonie commented Aug 23, 2017 • edited Loading

phildini commented Aug 23, 2017

abonie commented Aug 23, 2017

phildini commented Aug 23, 2017

phildini commented Aug 26, 2017

Choose a reason for hiding this comment

abonie commented Aug 27, 2017

abonie commented Aug 20, 2017 •

edited

Loading

abonie commented Aug 23, 2017 •

edited

Loading