|Listing installation units in p2 [message #721589]
||Fri, 02 September 2011 08:25
Originally posted by: |
We are trying to understand what is going on during our p2 installation
of a feature, more specifically we want to use the -list option and we
are trying it so (without having yet installed the feature of interest):
app.exe -noSplash -application org.eclipse.equinox.p2.director -list
where app.exe is the executable of our RCP program and
file://C:\Users\BlaBlub\something points to a directory that again
contains a directory eclipse which again contains a plugins and a
features sub directory. We always get no result and have checked that
this execution returns with result 0.
Could anyone shed some light on this?
Let me add that if we adapt the above command such that we perform the
actual feature installation as in
app.exe -noSplash -application org.eclipse.equinox.p2.director
-installIU somefeature.feature.group -repository
we get also a 0 result and we find them installed in our p2 repository.
We notice though, that the second step additionally creates a
(non-empty) file eclipse.ini.ignored: What is the purpose of this file?
(I could not find any helpful result hits during a google search for the
Thanks & Greetings from Bremen,
|Re: Listing installation units in p2 [message #723759 is a reply to message #723631]
||Fri, 09 September 2011 11:22
Originally posted by: |
On 2011-09-09 04:42, Brian de Alwis wrote:
>> Are you trying to explain to me that we cannot do those commands
>> (including feature installation and the like) with our just installed
>> RCP product (which is indeed a Windows-based application, no console
>> program)? It would seem not acceptable to have some additional
>> eclipsec.exe including its dependencies part of our own product setup
>> just to realize an p2 director installation/uninstallation. Or is it
>> just so, that the expected console result is written to some file?
> You can invoke these p2 apps, but their output goes to stdout. The
> eclipse.exe binary is a windowed application and won't show output sent
> to stdout.
> As Andrew Niefer put it
Thanks for the link, I have read this and the next link pointed to by
Andrew Niefer, but this does not answer the question to me, what the
requirements are for eclipsec.exe. Does it contain all the dependencies
it needs? Do we really have to copy this executable to out setup program
just to realize an p2 installation/uninstallation of a feature to an
existing RCP application? I hope we could do that with our RCP app (and
it seems already to work except for the missing output and for the
incomplete uninstall). Is there some documented way to create the
corresponding console version form of my own RCP application for this
>> eclipse/eclipsec is analogous to javaw/java. Eclipsec.exe is linked
>> as a console application so that the OS will allocate a console for
>> us. This executable does not automatically add any arguments, so you
>> still need to specify -console, -consoleLog, etc.
> Are you invoking the director from your program? Don't do this: look at
> the p2 APIs, See the following link for a starter:
I'm not sure what you mean by "invoking the director from your program".
This won't be done by the user nor from the running RCP application. We
don't integrate Eclipse update mechanisms via the UI. We only want to
install or uninstall an optional feature to an existing RCP installation
with p2 structure via our setup program using the console commands. We
expected that this can be done via the RCP.exe and in fact it seems to
work partially, as described above. What we didn't do was to add the part
to this command yet (I haven't found anywhere a reason why this is
necessary to do).
Thanks for your replies so far, even though I have a slightly bitter
aftertaste realizing the complexity of a seemingly simple thing such as
a feature install or uninstall to an existing p2 RCP application.
- Daniel Krügler
Powered by FUDForum
. Page generated in 0.21217 seconds