[tlbuild] dvi test?

[Richard Koch]

Paul,



> On Jan 27, 2022, at 6:35 PM, Paul Vojta <vojta at math.berkeley.edu> wrote:
> 
> Actually in my office I have an old Mac Mini 4,1 with MacOS Yosemite.
> 
> Would that be new enough to reproduce the bug?
> 
> Paul
> 
> 
> On Thu, Jan 27, 2022 at 03:07:43PM -0700, Karl Berry wrote:
>>    Subject: Re: [tlbuild] dvi test?
>> 
>> Does anyone have a Mac on which Paul (or I, if it has to be ...) could
>> ssh in and reproduce the problem? I don't think much more can be done
>> from a distance, unless Paul has a miraculous idea from the debug output.
>> 
>> (John already said that his Mac(s) aren't suitable for this.)
>> --thanks, karl.


As it happens, I have a machine here running Yosemite. 10.10.5. It contains TeX Live 2018, which I believe is the last version supporting Yosemite (except Mojca's binaries). It also had the old 2016 version of XQuartz.

I tried John's file with this setup and could not produce the bug. I also could produce no bug with my longer file with lots of illustrations. When using John's file, xdvi-xau wrote the following line several times in Terminal:

Warning: Could not find graphics file "13-backend-dvips.png"

This line did not occur with my file.

----------------

Next I tried to upgrade XQuartz, and surprisingly the latest version supports Yosemite. This allowed me to rerun all the tests above. They gave identical results with this latest XQuartz, and in particular no crashes.

However, my source has lots of illustrations, and a bug in previous versions of xdvi surfaced: at first illustrations appeared in correct spots, but then a buffer apparently wasn't erased and pieces of old illustrations appeared in new ones. This bug was fixed several years ago, but apparently after 2018.

-----------------

Then I tried something daring. I have a copy of xdvi-xaw produced this year for the upcoming TeX Live 2022. It will be universal, but I have the x86_64 version of that binary. So I copied it to the 2018 directory, replacing the original x86_64. I expected it to crash, but surprisingly it runs fine on Yosemite. I regret to report that it does not crash, either on John's file or on mine. However, I know that the new version of xdvi-xaw is running because the graphic bug I described earlier is gone with this version.

So I suspect your Yosemite machine will not show the bug, but this experiment produced some results which certainly surprised me.

Now I'm going to rerun my test of John's source on an Intel machine here running Monterey, just to confirm an earlier report that the bug occurs on both Arm and Intel machines. 

By the way, what is this "13backend-dvips.pnp" file. Are we certain that it is not the cause of the bug, since it is not present on this Yosemite machine.

Richard Koch