From: Christian Sternagel <c.sternagel@gmail.com>
Hi Makarius,
with RC3 I have the "offset-issue" again.
This time, and I think that is a first, this also impacts Ctrl-hover
however. Also the wmname trick as e.g. mentioned in
https://lists.cam.ac.uk/mailman/htdig/cl-isabelle-users/2013-December/msg00175.html
does not help anymore.
To be more precise (I'm on Fedora 20 x86_64 with Gnome): When I start
Isabelle/jEdit for the first time its window does not use the full
screen size (at this point menus and C-hover still works). So I usually
double-click on the title bar in order to make the window as big as the
screen. After that the mouse-pointer has an offset for all menus (i.e.,
I have to point somewhat lower and more to the right of the screen in
order to "hit" with the mouse). That alone would not be so bad. But also
now C-hover does not work anymore. More specifically, the word I point
at is still highlighted by a gray rectangle, as usual, but no popup
shows up.
With Isabelle2013-2 this issue does not occur (after using the wmname
trick). I did not check with RC1 or RC2 since I did not have them
readily available on my machine.
cheers
chris
From: Makarius <makarius@sketis.net>
On Thu, 14 Aug 2014, Christian Sternagel wrote:
with RC3 I have the "offset-issue" again.
This time, and I think that is a first, this also impacts Ctrl-hover however.
Also the wmname trick as e.g. mentioned inhttps://lists.cam.ac.uk/mailman/htdig/cl-isabelle-users/2013-December/msg00175.html
does not help anymore.
This looks like another episode of the cheap daily-soap "Oracle versus X11
Window manager forks".
With Isabelle2013-2 this issue does not occur (after using the wmname
trick). I did not check with RC1 or RC2 since I did not have them
readily available on my machine.
It should be just a matter of Java versions. The Isabelle2014-RC versions
use jdk-7u60 or jdk-7u65. In Isabelle2013-2 it was jdk-7u40.
You may also try different ones from http://isabelle.in.tum.de/components
using the normal Isabelle init_component setup that you probably know (as
explained in the "system" manual).
jdk-8u5 and jdk-8u11 might be worth trying, but note that the Java 8
releases are officially stable and de-facto still experimental.
Makarius
From: Makarius <makarius@sketis.net>
I have tried this briefly with the OpenJDK 7u65 variant that is presently
provided by Fedora 20, and it seems to work a bit better -- testing only
for 10min.
To switch the JDK version, you merely need to add something like this in
$ISABELLE_HOME_USER/etc/settings:
ISABELLE_JDK_HOME="/usr/lib/jvm/java-1.7.0"
Or any other directory that is produced by the java package installation.
That directory is supposed to contain a standard JAVA_HOME layout, but
this is only approximative here, due to the inevitable "packaging".
The above only works for "isabelle jedit" command-line use. The toplevel
Isabelle app takes its jdk from $ISABELLE_HOME/contrib/jdk via the
Isabelle.run shell script (which may be also changed, of course).
Makarius
From: Christian Sternagel <c.sternagel@gmail.com>
Dear Makarius,
thanks for the reminder (I already forgot that this is caused by the
JDK) and thanks for the hint.
I did not try your suggestion with OpenJDK yet since the issue
disappeared with RC4, but I'll keep it in mind for the future.
cheers
chris
From: Makarius <makarius@sketis.net>
Strange. I had tried it with Isabelle2014-RC4 myself, and doing this
again the problem is still there -- using the "latest" Fedora 20 Heisenbug
version, according to the update manager.
Makarius
From: Makarius <makarius@sketis.net>
The difference is probably due to wmname, which I did not manage to
install on Fedora.
If it works for you with that workaround, we can declare this thread
closed, presuming that it is somehow subsumed by the Oracle announcement
of jdk-7u67: "This release addresses a regression introduced in the most
recent security update." The officially documented regression is quite
different, but who knows how many unofficial changes there are.
Makarius
From: Christian Sternagel <c.sternagel@gmail.com>
Agreed. Anyway, I'll check whether it really is wmname. If I could only
remember what the "offending" wmname was ...
Last updated: Nov 21 2024 at 12:39 UTC