On Mon, 10 Apr 2017 08:52:52 -0400, Yuya Nishihara <y...@tcha.org> wrote:
On Sun, 09 Apr 2017 21:52:30 -0400, Matt Harbison wrote:
# HG changeset patch
# User Matt Harbison <matt_harbi...@yahoo.com>
# Date 1491785368 14400
# Sun Apr 09 20:49:28 2017 -0400
# Node ID 1de38b5ca1886aa41844dc7a80c1b92b65a8171e
# Parent 9259cf823690e4fcd34a4d2ecd57ced2060d2b3d
test-patchbomb: convert the pretendmail program from *.sh to *.py
Windows doesn't know how to execute *.sh, so tests were failing with:
abort: 'pretendmail.py' specified as email transport, but not in PATH
I was able to create a *.bat file to do the same thing, but neither the
*.py nor
*.bat were able to read from stdin[1] and write to stdout on Windows,
like `cat`
was doing in the shell script. The *.py script is infinitely more
readable.
It isn't possible to specify 'method = python `pwd`/*.py', because
everything
specified on the RHS ends up quoted, because of the space. Therefore,
$PATHEXT
is modified to include *.py, which causes a registry lookup and Windows
to
figure out that it should open *.py with python.exe. Unlike the python
installer, the MSYS installer doesn't add registry entries for *.sh, so
it isn't
possible to execute shell scripts directly.
Does it mean there's still a risk the test will fail? e.g. because .py is
associated with python3, or not associated at all.
It seems like a small risk to be associated with something other than
python, but not impossible. If it was associated with python3, that seems
OK since it's just reading stdin and writing to stdout (assuming this is
valid py3 code- I didn't try that).
I'm fine with disabling the tests though.
I think it's okay to disable the sendmail test at all on Windows because
sendmail is a Unix thing.
_______________________________________________
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel
