taverna-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michele Donato <mdon...@stanford.edu>
Subject Re: Taverna Workbench crashes on Mac OS
Date Thu, 09 Feb 2017 17:29:14 GMT
Hi Donald,

First of all, thanks for your reply.

One thing to notice, it is not a problem of the retina display, but it is related to the graphic
card. And, more specifically, Taverna does not start with the discrete card, but if I start
it with the integrated one, and then I switch, everything works correctly. This makes me think
that whichever component causes the crash it is not fundamental to the application.

In any case, I understand that solving this issue leads to a whole other sets of issues.

I will follow the development and try to see if the next versions work for me.



On Feb 9, 2017, at 02:33, Donal K. Fellows <donal.k.fellows@manchester.ac.uk<mailto:donal.k.fellows@manchester.ac.uk>>

On 07/02/2017 17:34, Michele Donato wrote:
Any idea how to fix this to have it work without the graphic card
switcharoo trick?

The problem is that it's a weird Java problem, and that's really out of
our league to fix. I've seen font-rendering problems in other simple
Java programs running on Retina displays, so I know it's not linked to
the Taverna code; there's at least a historic problem with the
differences between the handling of the glyph definitions and the
metrics for those glyphs when running on Retina displays, and maybe on
other HighDPI systems too. That there's a crash is bad, but it's hard to
pin it down to anything specific that we're doing wrong; we really don't
do very much over-clever with the GUI, or at least not in a way that
should produce a full process crash (as we just aren't using native code
that way).

The internals of font rendering are *very* complicated, in the “scare me
off every time I look” category (along with regular expression
evaluation engines), so the best that I think is possible at all is to
find an updated JRE and replace the internal version with something that
*might* be fixed. But without changing the major version of the JRE,
since that breaks other things (Taverna's critically dependent on both
security and classloader configuration, and those things *do* change
between major Java versions; one of the goals for Apache Taverna
Workbench is to stop having the worst of these problems, as they're
symptoms of sustainability problems).

I acknowledge that this is a singularly unhelpful message from your
perspective. Sorry about that…


View raw message