Home » Eclipse Projects » Equinox » what to expect from p2 sat solver?
| what to expect from p2 sat solver? [message #133075] |
Mon, 15 June 2009 11:20  |
 Stephan Herrmann
Messages: 1853
Registered: July 2009 |
Senior Member |
|
|
Hi,
I'm not sure what I may expect from p2 when I invoke
"Check for Updates" from the menu.
I have configured two repositories, the eclipse 3.5milestones
and one containing our own features. On the first repo I find
Eclipse SDK I20090605-1444 (aka RC4)
Eclipse SDK I20090609-1400 (oops, what milestone is that?)
In our repo I have a feature that is compatible with RC4 but not
with that unnamed newer milestone (due to a patch feature
requiring an exact feature version, see previous post).
Check for Updates selects the newest from both repos and
finds that they are not compatible. No surprise, wouldn't need
a constraint solver for that.
When I manually select the best matching versions installation
goes fine (yes, I migth mention that for a change, too ;-),
Is p2 supposed to find compatible versions, i.e., is the solver
traveling up-and-down the version history of items, or is it only
investigating the space, of WHICH elements to install, and only
ever try the newest version of each element?
best,
Stephan
PS: On the successful path I had difficulties with one specific
jar: during install p2 got stuck fetching pde.core, first attempt
timed out, then it looped over fetching the file, thinking a few
seconds and then fetching the file again, and again, and ...
After 10 or more attempts the download succeeded and the
update continued - pde.core is installed in the newest version.
Should we believe that downloading this particular file indeed
failed somewhere on the wire for roughly ten times (while hundrets
of other downloads from the same server worked ok), or does
this sound like a p2 (or file transfer) bug??
|
|
|
| Re: what to expect from p2 sat solver? [message #133121 is a reply to message #133075] |
Mon, 15 June 2009 15:17  |
Andrew Niefer
Messages: 990
Registered: July 2009 |
Senior Member |
|
|
I'll let someone else answer the SAT solver parts,
but I20090609-1400 was a build towards the final 3.5 platform.
Currently I20090611-1540 is under consideration to be promoted as the
final official 3.5 release.
For patches, the tooling currently only generates the patch against a
specific version because that was the semantics of the old update
patches. However, p2 is more flexible, and if you are willing to edit
the metadata by hand (or by xslt transform), you can change the patch to
match a range. See also the "Matching feature versions" thread.
http://www.eclipse.org/newsportal/article.php?id=6367&gr oup=eclipse.technology.equinox#6367
Stephan Herrmann wrote:
> Hi,
>
> I'm not sure what I may expect from p2 when I invoke
> "Check for Updates" from the menu.
>
> I have configured two repositories, the eclipse 3.5milestones
> and one containing our own features. On the first repo I find
> Eclipse SDK I20090605-1444 (aka RC4)
> Eclipse SDK I20090609-1400 (oops, what milestone is that?)
> In our repo I have a feature that is compatible with RC4 but not
> with that unnamed newer milestone (due to a patch feature
> requiring an exact feature version, see previous post).
>
> Check for Updates selects the newest from both repos and
> finds that they are not compatible. No surprise, wouldn't need
> a constraint solver for that.
> When I manually select the best matching versions installation
> goes fine (yes, I migth mention that for a change, too ;-),
>
>
> Is p2 supposed to find compatible versions, i.e., is the solver
> traveling up-and-down the version history of items, or is it only
> investigating the space, of WHICH elements to install, and only
> ever try the newest version of each element?
>
> best,
> Stephan
>
> PS: On the successful path I had difficulties with one specific
> jar: during install p2 got stuck fetching pde.core, first attempt
> timed out, then it looped over fetching the file, thinking a few
> seconds and then fetching the file again, and again, and ...
> After 10 or more attempts the download succeeded and the
> update continued - pde.core is installed in the newest version.
> Should we believe that downloading this particular file indeed
> failed somewhere on the wire for roughly ten times (while hundrets
> of other downloads from the same server worked ok), or does
> this sound like a p2 (or file transfer) bug??
>
|
|
|
Goto Forum:
Current Time: Sat Apr 20 00:23:06 GMT 2024
Powered by FUDForum. Page generated in 0.02411 seconds |
] 
Search
Help
Register
Login
Home
