Open merks opened 10 months ago
The only difference is the quotes around the
filter:=
's value. I don't think the header is invalid according to the OSGi specification; I think the header parsing utility is what's wrong.
I'm not sure about that. In the spec 3.2.7 Filter Syntax and 3.3.6 Bundle Requirements are probably the relevant sections but I cannot find anything about quotes there. Probably @tjwatson knows that best?
What should "we" do about this? ECF could add the quotes around the filter value, aQute ought to fix the parsing problem. and aries needs a copy of that fix.
Fixing this at ECF would probably the simplest. If this is really valid of course a fix in BND-tools would of course be good. And even if this is invalid, the error message should be a bit more precise.
Besides that, did you noticed any loss in functionality or is this just an annoying message?
I think it was just an annoyingly incorrect messages. Equinox itself has unquoted examples:
Then it is probably valid.
Did you test your standalone demo with the latest version of BND's OSGIHeader class? If this error occures there it is probably a good idea to report that at https://github.com/bndtools/bnd/issues. The BND devs are usually very well aware of the spec and will tell you if this is, for whatever reason, actually not a valid input.
I think it's the latest:
I think it's the latest:
Yes it is.
@HannesWell @jonahgraham @scottslewis
For lack of a better place to open an issue, opened it here...
Installing the Eierlegende Wollmilchsau 2023-09 RC1, this was logged:
No clue in the log about the bundle original of this problem, so I used remote debug to figure that out:
Of course this begs the question "why are examples are available to be installed"?
In any case, the problem arises here in aries parsing the header:
Here is a stand-alone program demonstrating the nature of the problem:
It prints out this:
The only difference is the quotes around the
filter:=
's value. I don't think the header is invalid according to the OSGi specification; I think the header parsing utility is what's wrong.Note that aries has a copy of that utility:
What should "we" do about this? ECF could add the quotes around the filter value, aQute ought to fix the parsing problem. and aries needs a copy of that fix.