Am Mittwoch, 19. April 2006 18:15 schrieb Joshua Nichols:
> Ivan Yosifov wrote:
> > Am Dienstag, 18. April 2006 19:36 schrieb Joshua Nichols:
> >> Ivan Yosifov wrote:
> >>> Hello,
> >>>
> >>> Happens while emergeing eclipse:
> >>>
> >>> home ~ # emerge -1 eclipse-sdk
> >>> Calculating dependencies... done!
> >>>
> >>>>>> Emerging (1 of 1) dev-util/eclipse-sdk-3.1.2 to /
> >>>>>> checksums files ....
> >>>>>> checksums src_uri ;-) eclipse-sourceBuild-srcIncluded-3.1.2.zip
> >>>
> >>>  * Checking for sufficient physical RAM
> >>>  * Checking for bad CFLAGS
> >>>  * Found offending option -fomit-frame-pointer in your CFLAGS
> >>> ....
> >>>  * Tip: use equery depgraph "=eclipse-sdk-3.1.2" to list all
> >>> dependencies.
> >>>
> >>> /usr/local/portage/migration/migration/eclass/java-utils.eclass: line
> >>> 97: [: -ge: unary operator expected
> >>> Detected a JDK < 1.5.0
> >>> Detected a JDK >= 1.4.2
> >>>
> >>>>>> Unpacking source...
> >>>
> >>> /usr/local/portage/migration/migration/ is where I've put the migration
> >>> overlay revision 2114. Bash is 3.1_p17. Is this an eclass bug or
> >>> something else ?
> >>
> >> Something else, I think. I've been working on the eclipse ebuilds (while
> >> using migration-overlay, obviously), and hadn't noticed any problems.
> >>
> >> It might have something to do with the ebuild itself. Someone went and
> >> committed a bump to 3.1.2 without consulting with us Java folks, so I'm
> >> not sure what condition it is. Expect to see a new revision in a day or
> >> so...
> >>
> >> Josh
> >
> > I am getting the same with the new eclipse-sdk-3.1.2-r1. I added some
> > debugging output to the eclass in java-utils_is-vm-version-ge():
> >
> > --- ./java-utils.eclass 2006-04-19 10:58:43.000000000 +0300
> > +++ /usr/local/portage/migration/migration/eclass/java-utils.eclass
> > 2006-04-19 11:00:25.000000000 +0300
> > @@ -91,6 +91,12 @@
> >         local vm_patch=$(echo ${vm_version} | cut -d. -f3)
> >         local vm_extra=$(echo ${vm_version} | cut -d. -f4)
> >
> > +       echo ${vm_major}
> > +       echo ${vm_minor}
> > +       echo ${user_major}
> > +       echo ${user_minor}
> > +       echo ${vm_patch}
> > +       echo ${user_patch}
> >         if [ ${vm_major} -ge ${user_major} ] && [ ${vm_minor} -gt
> > ${user_minor} ] ; then
> >                 echo "Detected a JDK >= ${user_version}"
> >                 return 0
> >
> > and the output is:
> >
> >  * Tip: use equery depgraph "=eclipse-sdk-3.1.2-r1" to list all
> > dependencies.
> >
> > 1
> > 5
> > 1
> > 5
> >
> > 0
> > /usr/local/portage/migration/migration/eclass/java-utils.eclass: line
> > 103: [: -ge: unary operator expected
> > Detected a JDK < 1.5.0
> > 1
> > 5
> > 1
> > 4
> >
> > 2
> > Detected a JDK >= 1.4.2
> >
> >>>> Unpacking source...
> >
> > ,meaning that vm_patch is ending up empty and that seems to confuse -ge
> > conditionals using it.
>
> I'll have to look into this, but it probably just needs to sanity
> checking for the vm_extra and company.
>
> Aside from that, what is the original of the ebuild for eclipse that you
> are using?  Also, which VM are you using? I haven't been able to quite
> reproduce this problem...
>
> Josh

I am using the eclipse-sdk-3.1.2-r1 ebuild from the main tree, attached here 
for reference. I have sun-jdk 1.4.2.10-r13 and 1.5.0.06-r13 installed. The 
problem with vm_patch seems to be that 

java-config -f | sed -e "s/.*-\([0-9.]\+\).*/\1/" ( in 
java-utils_get-vm-version )

returns just "1.5" and so 

vm_patch=$(echo ${vm_version} | cut -d. -f3) ( in 
java-utils_is-vm-version-ge ) ends up being an empty string. There is no 
third field to cut.

-- 
Cheers,
Ivan Yosifov.