[Numpy-discussion] Initial Hg experiences on Windows

David Cournapeau cournape@gmail....
Sun Jan 6 10:51:46 CST 2008

On Jan 7, 2008 1:32 AM, Bill Baxter <wbaxter@gmail.com> wrote:
> I've been playing around with Hg on windows for an hour or so now.  My
> overall impression is that the installation process isn't quite there
> yet.
> The basic binary installer goes very smoothly, and after that I was
> able to open up a prompt and type hg commands right away.  But going
> through the tutorial I ran into some issues:
> *) After running the binary installer, apparently you're supposed to
> go edit some .ini file to specify your username.  It seems it will
> work ok even if you don't set your username, but since it is
> apparently highly recommended, the installer just should ask you as
> part of the install process.
> *) Despite all this talk about great merging capabilities, for Windows
> there is *no* merge functionality installed by default.  The merge
> page (http://www.selenic.com/mercurial/wiki/index.cgi/MergeProgram)
> explains how to get merging working on Windows.  But what it gives you
> is a confusing list of alternatives without offering any guidance as
> to what the trade-offs are.  It mentions "batteries included" binary
> distributions as one solution without giving any link.  It offers an
> hgmerge.py script as a solution but then is wishy-washy about how you
> specify it in the Mercurial.ini file (just "hgmerge.py" or "python
> \path\to\hgmerge.py"?).  The hgmerge.py script itself relies on
> pywin32 for some of its functionality, but that is not mentioned
> anywhere (well, it is now because I edited the wiki...)  Without that
> it fails to find some installed merge tools but catches the import
> exceptions and reports no error.   The page also fails to give a clear
> recommendation as to which merge tool to use, and they are *not* all
> created equal.  It managed to find DiffMerge already installed on my
> system, but DiffMerge does a terrible job of automatic merging.  It
> couldn't properly merge two lines added to different parts of a
> hello_world.c.
> *) After failing to run any merge program, I got the hgmerge.py script
> installed and tried to redo the merge.  No dice "error: uncomitted
> merge pending".  But I didn't merge anything yet!  A hint about how to
> rerun or undo the failed merge would have been a little more user
> friendly.  I fortunately stumbled across a web page telling me about
> hg rollback, and that seemed to do the trick.
> So I think the installation on Windows is a ways away still from what
> I get from installing TortoiseSVN, particularly the merge thing.
> Hg/windows really really needs to come bundled with at least a
> rudimentary merge program that's all configured properly.

I don't think that any DVCS is where svn is in this regard; mercurial
does not seem to be, bzr is not either, and git certainly is not.

As far as command line is concerned, I know for a fact that the
official bzr installer does work on windows (I use bzr on windows, in
particular for my work related to scons integration with distutils).

I think that at least wrt windows support, bzr is today better than
hg, if only for the fact that some bzr developers are primely windows
developers, with the "windows culture".


More information about the Numpy-discussion mailing list