std.math: deprecate backwards compatibility with Phobos1

[9il]

is____ functions in std.math

Depends on dlang/dmd#4090

[quickfur]

Looks like there's a merge conflict; please rebase.

[quickfur]

Auto-merge toggled on

[9il]

Test failed: expected rc == 0, exited with rc == 1

What does it mean?

[IgorStepanov]

@9il
rc == return code. You need to scroll above to see the actual error.
Win32 fail

==============================

Test failed: 

expected:

----

fail_compilation\fail80_m64.d(28): Error: cannot implicitly convert expression ("progress_rem") of type string to ulong

fail_compilation\fail80_m64.d(29): Error: cannot implicitly convert expression ("redo") of type string to ulong

----

actual:

----

C:\cygwin\home\Administrator\sandbox\d\d-tester\client\pull-1230416-Win_32\dmd\src\..\..\phobos\std\format.d(33): Deprecation: alias std.math.isnan is deprecated - Phobos1 math functions are deprecated, use isNaN

fail_compilation\fail80_m64.d(28): Error: cannot implicitly convert expression ("progress_rem") of type string to ulong

fail_compilation\fail80_m64.d(29): Error: cannot implicitly convert expression ("redo") of type string to ulong

----

Win64 fail

Z:\Administrator\sandbox\d\d-tester\client\pull-1230429-Win_64\dmd\src\..\..\phobos\std\format.d(33): Deprecation: alias std.math.isnan is deprecated - Phobos1 math functions are deprecated, use isNaN



compilable\diag10768.d(36): Deprecation: implicitly overriding base class method diag10768.Frop.frop with diag10768.Foo.frop deprecated; add 'override' attribute







==============================

Test failed: 

expected:

----

compilable\diag10768.d(36): Deprecation: implicitly overriding base class method diag10768.Frop.frop with diag10768.Foo.frop deprecated; add 'override' attribute

----

actual:

----

Z:\Administrator\sandbox\d\d-tester\client\pull-1230429-Win_64\dmd\src\..\..\phobos\std\format.d(33): Deprecation: alias std.math.isnan is deprecated - Phobos1 math functions are deprecated, use isNaN

compilable\diag10768.d(36): Deprecation: implicitly overriding base class method diag10768.Frop.frop with diag10768.Foo.frop deprecated; add 'override' attribute

----

[quickfur]

Looks like one of the tests is failing, probably the output doesn't exactly match the expected output.

[quickfur]

P.S. Ah, looks like you have to update some of the dmd test files to either use the new function names, or update the expected output to include the new deprecation messages.

[quickfur]

Auto-merge toggled off

[9il]

Ok, probably I understand. I need to rewrite Phobos1 tests =)

[quickfur]

Hmm, the autotester is passing now. Was there a corresponding dmd PR that got merged?

[9il]

isnan was called from std.format, not from test: See 9il@cb8d75e .
dlang/dmd#4090 it is compiles or not compiles test, and it is compiles with deprecated feature.

[quickfur]

Ah, makes sense. Thanks! LGTM.

[quickfur]

Auto-merge toggled on

[WalterBright]

These name changes have to stop. They're the poster boy for churn giving an "illusion of progress".

[mihails-strasuns]

Never underestimate learning curve damage of inconsistent naming.

[WalterBright]

Never underestimate learning curve damage of inconsistent naming.

I don't buy that. There's nothing "intuitive" or "consistent" with isNaN instead of isnan. (Googling it shows that isnan, isNaN, ISNAN, is-nan are common spellings in various languages.) There is, however, considerable ongoing damage with "my working code from 1 year ago won't compile anymore because of gratuitous incompatible changes."

We ought to be working on things that matter, like making the documentation for isnan better, than just more wretched bikeshedding and breaking.

[yebblies]

I don't buy that. There's nothing "intuitive" or "consistent" with isNaN instead of isnan.

Of course there is, because we have a naming style and isnan doesn't follow it.

[WalterBright]

Of course there is, because we have a naming style and isnan doesn't follow it.

Then it should have been isNan. In any case, it's been isnan for 14 years. Breaking long standing, working, correct code like this for no good reason just annoys people. It wastes our time and every user's time. It must stop. I'd be shocked if isnan confused anybody.

[mihails-strasuns]

In any case, it's been isnan for 14 years

It was isNaN for many years. isnan as simply an alias hanging around undocumented and frustrating newcomers when they notice is used (is it different from "official" isNaN? why it is used?). All this PR does is finally marking it as deprecated to give clear message that it is legacy. It breaks no code at all and clarifies potential confusion. We need more cleanup like that.