Test Bug Report and a Patch on pytest'vspec/signatures.vim failed'

[gyx47]

Description:

There is a problem with the test code of the vim-jedi 0.11.2-3 package. The package build failed under both the riscv64 and x86-64 architectures. According to the log output, the signature test file reported an error when entering the test phase.

<buffer=5>
call jedi#clear_call_signatures()
        Last set from ~/vim-jedi/jedi-vim-0.11.2/autoload/jedi.vim line 604
if exists('b:_jedi_orig_updatetime') |   let &updatetime = b:_jedi_orig_updatetime |   unlet b:_jedi_orig_updatetime | endif
        Last set from ~/vim-jedi/jedi-vim-0.11.2/autoload/jedi.vim line 608
ok 2 - signatures multiple buffers
2 buffers wiped out
ok 3 - signatures simple after CursorHoldI with only parenthesis
ok 4 - signatures highlights correct argument
ok 5 - signatures no signature
ok 6 - signatures signatures disabled
staticmethod(f: Callable[..., Any])
foo(a, b)
ok 7 - signatures command line simple
not ok 8 - signatures command line truncation
 Expected Signature() == "\n".funcname."(arg1, …)" at line 15
     Actual value: "\n"
     Expected value: "\naaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(arg1, …)"
ok 9 - signatures command line no signature

FAILED test_integration.py::test_integration[vspec/signatures.vim] - Failed: vspec/signatures.vim failed:
======================================= 1 failed, 7 passed, 3 warnings in 6.73s ============================

• package version(s):vim-jedi 0.11.2-3，Attached is a patch:python-3.13-accept-test-results.patch
• config and/or log files:

vim-jedi-0.11.2-3-riscv64-prepare.log

vim-jedi-0.11.2-3-riscv64-check.log

My operating environment is an x86-64 environment virtual machine of the arch architecture of Windows WSL. The process is as follows:

1. I cloned your project and installed the dependencies
2. Then execute py.test
3. Wait for about 7 seconds and find an error

Patch

I modified the test file. Here are my ideas for modification：

This is the relevant code for the command line truncation error in the eighth test of /test/vsepr/sinagature.vim

it 'command line truncation'
        let g:jedi#show_call_signatures = 2
        call jedi#configure_call_signatures()

        function! Signature()
            redir => msg
            python3 jedi_vim.show_call_signatures()
            redir END
            return msg
        endfunction

        let funcname = repeat('a', &columns - (30 + (&ruler ? 18 : 0)))
        put = 'def '.funcname.'(arg1, arg2, arg3, a, b, c):'
        put = '    pass'
        execute "normal o\<BS>".funcname."( "
        Expect Signature() == "\n".funcname."(arg1, …)"

        exe 'normal sarg1, '
        Expect Signature() == "\n".funcname."(…, arg2, …)"

        exe 'normal sarg2, arg3, '
        Expect Signature() == "\n".funcname."(…, a, b, c)"

        exe 'normal sa, b, '
        Expect Signature() == "\n".funcname."(…, c)"

        g/^/d
        put = 'def '.funcname.'('.repeat('b', 20).', arg2):'
        put = '    pass'
        execute "normal o\<BS>".funcname."( "
        Expect Signature() == "\n".funcname."(…)"
    end

Here is the relevant code for the show_call_signatures function called by this test

def show_call_signatures(signatures=()):
    if int(vim_eval("has('conceal') && g:jedi#show_call_signatures")) == 0:
        return

    #We need to clear the signatures before we calculate them again. The
    #reason for this is that call signatures are unfortunately written to the
    #buffer.
    clear_call_signatures()
    if signatures == ():
        signatures = get_script().get_signatures(*get_pos())

    if not signatures:
        return

    if int(vim_eval("g:jedi#show_call_signatures")) == 2:
        return cmdline_call_signatures(signatures)
…

After debugging, I found that according to the original test code, the process should have entered the fourth if function to execute g:jedi#show_call_signatures = 2, but in fact it returned in the third if, and no content was output, resulting in the return of \n. The reason is that signatures = get_script().get_signatures(*get_pos()) is an empty list And the signature cannot be read. I think it is related to <BS> returning to the end of the previous line and directly inputting, and it is impossible to return the signature according to the independently input function name.

I deleted the <BS> in the code in the patch, and then performed the same operation as above,

let funcname = repeat('a', &columns - (30 + (&ruler ? 18 : 0)))
        put = 'def '.funcname.'(arg1, arg2, arg3, a, b, c):'
        put = '    pass'
        execute "normal o".funcname."( "
        Expect Signature() == "\n".funcname."(arg1, …)"

        exe 'normal sarg1, '
        Expect Signature() == "\n".funcname."(…, arg2, …)"

        exe 'normal sarg2, arg3, '
        Expect Signature() == "\n".funcname."(…, a, b, c)"

        exe 'normal sa, b, '
        Expect Signature() == "\n".funcname."(…, c)"

        g/^/d
        put = 'def '.funcname.'('.repeat('b', 20).', arg2):'
        put = '    pass'
        execute "normal o".funcname."( "
        Expect Signature() == "\n".funcname."(…)"

and found that the test passed

Do you think this is a good change? If you think it is, I will open a PR.

[eli-schwartz]

I modified the test file. Here are my ideas for modification：

Why don't you submit this as a PR instead? :)