[PATCH 3 of 3 in crew] run-tests: use correct python safely under --jobs

[Adrian Buehlmann]

On 2012-12-12 02:35, Siddharth Agarwal wrote:
> Unfortunately it seems the trick of copying the python executable into 
> the hg directory doesn't work on Windows, because Windows Python depends 
> on other files in its own directory to work.

This is not entirely true, as at least a globally installed Python on
Windows will have the required dll's installed in a system directory and
should thus find them (and its other required files).

At least it does find them here (Windows 7 x64).

The tests pass here in MSYS using fe5a41144982 with:

$ PATH="/mingw/bin:/bin:`pwd`/..:/c/python:/c/python/scripts" python
run-tests.py --local

Two tests are currently failing here:

--- C:\Users\adi\hgrepos\hg-main\tests\test-install.t
+++ C:\Users\adi\hgrepos\hg-main\tests\test-install.t.err
@@ -1,7 +1,7 @@
 hg debuginstall
   $ hg debuginstall
   checking encoding (ascii)...
-  checking Python lib (*lib*)... (glob)
+  checking Python lib (C:\Python27_x64\Lib)...
   checking installed modules (*mercurial)... (glob)
   checking templates (*mercurial?templates)... (glob)
   checking commit editor...
@@ -11,7 +11,7 @@
 hg debuginstall with no username
   $ HGUSER= hg debuginstall
   checking encoding (ascii)...
-  checking Python lib (*lib*)... (glob)
+  checking Python lib (C:\Python27_x64\Lib)...
   checking installed modules (*mercurial)... (glob)
   checking templates (*mercurial?templates)... (glob)
   checking commit editor...

--- C:\Users\adi\hgrepos\hg-main\tests\test-rebase-pull.t
+++ C:\Users\adi\hgrepos\hg-main\tests\test-rebase-pull.t.err
@@ -144,7 +144,7 @@
   adding L1
   created new head
   $ hg pull --rev tip --rebase
-  pulling from $TESTTMP/a
+  pulling from $TESTTMP\a
   searching for changes
   adding changesets
   adding manifests

After the test, I see that there is a python.exe in the repo root.

And Microsoft's Windows "firewall" popped up alerting me that this
python.exe newly wants to start listening on some ports (which I
acknowledged).

> Mads suggested we make sure
> os.path.join(dirname(sys.executable), "python.exe") existed, and then 
> add that directory to the path.

I haven't investigated enough to comment on that.

But copying the python.exe feels indeed a bit strange on first thought here.