[superseded] fix #14142: no more clash with: import os + use of existsDir/dirExists/existsFile/fileExists/findExe in config.nims

[timotheecour]

/cc @Araq

• fix import os + use of existsDir/dirExists/existsFile/fileExists/findExe in config.nims causes "ambiguous call' error #14142

future work

• implicitly exported symols are not good, in particular when they differ from standard nim c behavior
• user should import os for these and it'd work via vmops
• we should be able to do this without much breakage by introducing a flag --nimsimports:off with semantics:
  --nimsimports:off # doesn't import anything, user has to import relevant module (eg os); will be default in some future tag 1.3.X
  --nimsimports:on # default until 1.3.x, gives a new warning while it's on

[Araq]

I spend effort on this error message, why remove it?

[timotheecour]

because the problems in #14142 are real and outweigh the benefits of having a nicer error message.

In following example, pkg/foo.nim (eg 3rd party code) can't be imported in some nimscript file because it'd give redefinition errors, same as #14142. Valid nim code should, as much as possible, be usable in nimscript without having to be modified with when defined(nimscript) sections.

example

nim e main.nims

# in main.nims (eg user config.nims or any other nims file)
import pkg/foo

# in pkg/foo.nim
import os
echo existsFile("/tmp/z01.txt") # Error: redefinition

{.error.} is the wrong pragma

We should instead use the new {.disabled.} pragma introduced in nim-lang/RFCs#213; if/when it's introduced, we can consider using:
{.pragma: noNimScript, disabled: "this proc is not available on the NimScript target".}

note: nimErrorProcCanHaveBody, introduced in #9665 (if n.sons[bodyPos].kind != nkEmpty and sfError notin s.flags: ...) improved {.error.} a bit but doesn't address the main problem, which causes #14142

[timotheecour]

superseded by #14658 , will be closed when #14658 is merged

[Araq]

The error message must stay.

[timotheecour]

I don't understand why producing an error here is so important; what is your alternative proposal to #14658 to fix #14179, #14142 and the regression introduced in 3d2459b that makes CI 1.4X times slower (or more generally makes nim slower for any command, which especially matters for short lived commands like nim dump, inim, tests etc)

[timotheecour]

superseded by #14658