Re: master 84e4bc6219a 1/2: Add input methods for Urdu, Pashto and Sindhi (bug#66470)

[Rahguzar]

Actually I see what is happening now. `cus-test-deps` is checking if it
can load the files defining the custom variables. None of the files in
lisp/leim/quail are on the load path. The cases where require succeeds
it is because some other file with the same file-name-non-directory is
on the load path e.g. for indian.el it is,

lisp/language/indian.el

which does not define any custom variable.

I think the problem is with the test.

P.S. Both lisp/leim/quail/indian.el and lisp/language/indian.el provide
the feature indian which doesn't seem like the right thing.

Rahguzar <rahguzar@zohomail.eu> writes:

> Hi Michael,
>
> Michael Albinus <michael.albinus@gmx.de> writes:
>
>> Rahguzar <rahguzar@zohomail.eu> writes:
>>
>>> Hi Michael,
>>
>> Hi Rahguzar,
>>
>>> So I wonder if the test is encouraging the correct behavior. I think
>>> that seeing the `pakistan-urdu-input` subgroup would be useful for
>>> people using `urdu-custom` input method but perhaps for majority of
>>> people we shouldn't cause pakistan.el to load just because they wanted
>>> to customize a quail option.
>>>
>>> Maybe it is better to just remove `quail` as the parent group of
>>> `pakistan-urdu-input`?
>>
>> No idea. Note, that there is also lisp/leim/quail/indian.el. It declares
>> defgroup `tamil-input', and it doesn't show an error in the test.
>>
>> I've tried to replace the parent group of `pakistan-urdu-input' to be
>> `leim' instead of `quail'. But the error about the missing pakistan.el
>> persists, just using `leim' instead of `quail' in the error message.
>
> Poking around `cus-test.el` a little I think this is a load path problem
> and unrelated to the defgroup form.
>
> I can run
>
> src/emacs -batch --eval="(require 'indian)"
>
> in a shell and it exists without any error.
>
> But
>
> src/emacs -batch --eval="(require 'pakistan)"
>
> results in
>
> Error: file-missing ("Cannot open load file" "No such file or directory" 
> "pakistan")
>   mapbacktrace(#f(compiled-function (evald func args flags) #<bytecode 
> -0x188ce6540514dc1>))
>   debug-early-backtrace()
>   debug-early(error (file-missing "Cannot open load file" "No such file or 
> directory" "pakistan"))
>   require(pakistan)
>   eval((require 'pakistan) t)
>   command-line-1(("--eval=(require 'pakistan)"))
>   command-line()
>   normal-top-level()
> Cannot open load file: No such file or directory, pakistan
>
> Similarly starting with
>
> src/emacs -Q
>
> M-: (require 'indian) RET
>
> works
>
> but
>
> M-: (require 'pakistan) RET
>
> doesn't.
>
> Neither do any of,
>
> M-: (require 'ipa) RET
> M-: (require 'uni-input) RET
>
> If fact there are 10 files in lisp/leim/quail with provide forms.
>
> Out of these requiring one philippine, indonesian, indian, misc-lang and
> emoji succeeds.
>
> While requiring one of uni-input, rfc1345, compose, pakistan or hangul
> leads to an error.
>
> I don't know why this is the case.
>
>> Best regards, Michael.
>
> Rahguzar