[libvirt] Got time to try out the libvirt 0.8.6 win32 package?

[Justin Clift]

On 14/12/2010, at 12:29 AM, Matthias Bolte wrote:
> I tested it and it works with ESX and QEMU over TLS. Here are some comments:
> 
> 1) I wonder why you never pushed this patch:
> https://www.redhat.com/archives/libvir-list/2010-November/msg00761.html

Heh, because it never got ACK'd. :)

Was mostly waiting for the TLS stuff to work well enough to be practical, which
I reckon your recent patches for %APPDATA% location should have done.

So, going to make a v2 of that patch, mentioning ESX support and location of the
TLS certificates.


> 2) libcurl license is missing in the installer.

Ugh, you're right.  Even after doing the work to locate and copy it into the installer
source tree, missed getting it into the package itself. :/

That's now fixed, and I've just pushed a new version of the 0.8.6 installer (using
exact same name) to the libvirt site.


> 3) There is no readme.txt about where to put TLS certs to make qemu+tls:// work.

Yeah, was thinking that should go on the Windows page.

But, now you mention it, having a readme or similar installed with the package
itself is probably a good idea.  I'll create a new version of the installer, including
it.


> 4) The installer remembers the install location. Therefore, it suggest
> to install in the same location as the previous installer for libvirt
> 0.8.5. But it doesn't check if a previous version is already installed
> and doesn't suggest to uninstall it first. This results in libvirt
> 0.8.6 being installed over 0.8.5. The installer overwrites existing
> files without notice.

Ouch, good point.  Now sure how to work around this atm, though in theory
it should be a matter of doing what you mention... detect installation already
in existence, offer to uninstall.  Also refusing to install over the same location
as an existing installer in case the user doesn't want to install.

For now, probably worth putting a note of the problem on the to-be-done
Windows page, and mentioning it in the to-be-done readme file.

But, it really needs to be fixed in a subsequent release. :)