[cfe-dev] shell-preserves-root is (*almost) always true...kill it?

[Yaron Keren]

OK, done finally in r242329.
With MSYS bash, one test still fails due to the / expansion and two other
tests fail because MSYS env does not pass TZ= to the executed program, see
discussion at

  https://github.com/Alexpux/MSYS2-packages/issues/294.

Hopefully MSYS developers will provide a workaorund this or we'll disable
the tests for MSYS.

Yet another option for LLVM development with MSYS2 is to use a bare-bones
env.cpp I wrote, good enough to pass LLVM tests but of course not a real
replacement for env:

#include <process.h>
#include <stdlib.h>
#include <string.h>

int main(int argc, char *argv[]) {
  for (unsigned i = 1; i < argc; ++i) {
    if (!strchr(argv[i], '='))
      _execvp(argv[i], argv + i);
    _putenv(argv[i]);
  }
}



2015-07-15 18:13 GMT+03:00 NAKAMURA Takumi <geek4civic at gmail.com>:

> It has been not maintained since r179283. and I haven't build mingw tree
> with autoconf. I don't object to drop shell-preservers-root.


On Wed, Jul 15, 2015 at 11:20 PM Yaron Keren <yaron.keren at gmail.com> wrote:

> Since r179283, internal shell is default on windows (*unless use_lit_shell
> = 0), so execute_external = not windows and
>
> # Exclude MSYS due to transforming '/' to 'X:/mingwroot/'.
> if not platform.system() in ['Windows'] or not execute_external:
>     config.available_features.add('shell-preserves-root')
>
> is really "if not windows or (not not windows)" = "if true" making
> shell-preserves-root always true on Windows (unless use_lit_shell = 0).
>
> No tests that require shell-preserves-root are actually disabled now on
> WIndows + MSYS2. In this configuration, just one test fails (which really
> require shell-preserves-root): darwin-sdkroot.c. This test has no purpose
> on windows anyhow and may be disabled. All other tests
> with shell-preserves-root do pass on MSYS2 bash.
>
> Does shell-preserves-root serves any purpose now?
> Any reason to keep it alive?
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20150715/94ae3ddf/attachment.html>