Hello,
I've made (as a test) an Analysis 2.0.7 release (with the patches that
have occurred so far) with Tcl/Tk 8.4 for 32-bit and 64-bit Linux,
available in the usual spot at:
http://www.ccpn.ac.uk/ccpn/software/downloads-v2
So I would recommend anyone (on Linux) who is having trouble with Tcl/Tk
8.5 to try this out and see if it helps (hopefully it will, otherwise we
are back to a puzzle).
Wayne
On Wed, 15 Apr 2009, Wayne Boucher wrote:
> Hello,
>
> Following that thread through some more, it seems that they're blaming some
> X11-type library called xcb. Apparently we're supposed to be impressed that
> the xcb developers "have used the Z formal method to prove the code correct
> in some areas". But it doesn't help if it's causing crashes elsewhere. On
> the Linux box where I built the 64-bit version of Analysis it doesn't have
> xcb but it does on the Linux box where I built the 32-bit version.
>
> The people in that thread suggest that one builds libX11 with xcb disabled
> (their comment is "libxcb is still not mature enough in my opinion. I think
> that it is definitely getting there, but not there yet."). Needless to say
> that is not very practical, since who on earth builds their own X11
> libraries. It's possible that this issue has been sorted with more recent
> releases of xcb so you could try and update that library, assuming that xcb
> really is the cause of the problem (and it's not totally obvious to me that
> the xcb problem is what is causing this Tk8.5 problem, this could all be a
> red herring).
>
> If this really is an xcb issue then I suspect the best way around this is to
> disable the aliased fonts in Tk8.5 or just use Tk8.4. People who compile
> Tk8.5 themselves can do this. Peope who have pre-compiled versions of
> Analysis cannot. I guess I ought to provide a Tk8.4 pre-compiled version of
> Analysis at least for Linux and see if this sorts people out.
>
> (Has anyone had this problem on OSX? That thread I am referring to is for
> OSX, so it sounds like people can have that issue there as well.)
>
> Wayne
>
> On Wed, 15 Apr 2009, Wayne Boucher wrote:
>
>> Hello,
>>
>> (Repeating a private email I sent on this very subject earlier today,
>> slightly edited.)
>>
>> There are not many examples of this kind of crash in google but I found one
>> page which sounds relevant (from just a week ago!):
>>
>> http://lists.macosforge.org/pipermail/xquartz-dev/2009-April/002250.html
>>
>> Now the implication there is that it's all down to some new library being
>> used (for anti-aliased fonts) in Tk8.5, which is why people won't have seen
>> it until recent Analysis releases (since we used to use 8.3 and 8.4). We
>> use the function he mentions on that webpage, Tk_DrawChars, to draw text
>> for assignments in the spectrum windows.
>>
>> One option is to turn this library (and so I assume font aliasing) off (but
>> note that that has to be done before compilation of Tk), and this is not
>> really very satisfactory because it would also affect the Analysis menus.
>> (And some people think that aliased fonts look good, although it doesn't
>> really do much for me.)
>>
>> The chap on that page claims it only happens in a 1-bit deep pixmap and we
>> don't (explicitly) create any 1-bit deep pixmaps. We use a pixmap which is
>> supposed to be of the same depth as the screen, and I would have thought
>> that was 8-bit for all screens these days. So my guess is that there is
>> some other subtle bug to do with anti-aliased fonts that these people
>> haven't quite sorted out (or even discovered) yet. But the odds are that
>> it's some kind of memory trample and it's possible that it's the Analysis
>> code that is causing this, but I would have thought not since it's only
>> come up with Tk8.5 (I assume, anyway; if anyone has seen this otherwise
>> then let me know).
>>
>> This function is only used in the Tk mode graphics. So in the OpenGL mode
>> there should not be a problem (except for the structure viewer and molecule
>> viewer, which both use Tk). But note that for Linux pre-compiled releases
>> the OpenGL mode is not enabled (but could be). (For OSX it is enabled and
>> I'd definitely recommend that people use it.)
>>
>> Wayne
>>
>> On Wed, 15 Apr 2009, Dr. Pryank Patel wrote:
>>
>>>
>>> Hi all,
>>>
>>> I know other people have had this issue before with Analysis 2.0.5 and
>>> 2.0.6 from looking at the email archives (Subject "X error"). I did not
>>> have this issue before, but I am experiencing this issue now, and has
>>> happened 3 times today:
>>>
>>> X Error of failed request: BadIDChoice (invalid resource ID chosen for
>>> this connection)
>>> Major opcode of failed request: 155 (RENDER)
>>> Minor opcode of failed request: 4 (RenderCreatePicture)
>>> Resource id in failed request: 0x3abd686
>>> Serial number of failed request: 36520783
>>> Current serial number in output stream: 36520795
>>>
>>>
>>> Any ideas? I am using the Tk graphics handler (I usually use openGL but am
>>> having problems with that as well). My OS is openSuse 11.1.
>>>
>>> When I use openGL, I frequently experience a total system freeze which
>>> requires me to restart my machine (I am using the latest graphics
>>> drivers).
>>>
>>> PS: is there any documentation on using the Backup system in Analysis? ie:
>>> how to load up a project from automated backup.
>>>
>>>
>>
>
|