Hi Bruce,

On Sun, Aug 9, 2026 at 12:43 AM <bruce@tomo.co.nz> wrote:

The guidance in texmf/web2c/texmf.cnf suggests 8000000 is the maximum, so I stopped there for a while, but then I tried 9000000.

And success!

... 

So if the limit is not 8000000, is there a limit?

Or are there possible problems with going over 8000000?


Possibly some configuration defaults for pdfTeX in TeXLive were changed without updating the comments in texmf.cnf.

I think maybe it's time to switch from pdfTeX to LuaTeX, which uses dynamic memory allocation (so the limit is the amount of installed RAM). It's a simple configuration change briefly described in the annex of the Thbook (search for luatex). The downside is that LuaTeX is a bit slower than pdfTeX, but it would be interesting to hear about your experience with it.
 

Might needing so much memory allocated suggest some problem with my dataset?

 

Curious that ‘maps off’ requires more memory than ‘maps on’ (this is a 70 km cave but there are scraps drawn for only 15 km)


That's weird. I could perhaps investigate it on your dataset.

Cheers
Martin 

Any insights appreciated.

 

Bruce

 

 

From: Therion <therion-bounces@speleo.sk> On Behalf Of Martin Budaj
Sent: Thursday, 21 May 2026 05:29
To: List for Therion users <therion@speleo.sk>
Subject: Re: [Therion] th_formdef.tex runaway text error

 

Hi,

 

after testing Bruce's dataset, it seems TeX really runs out of memory in this case. 

 

You can increase it in the file texmf/web2c/texmf.cnf. Set

 

extra_mem_top = 4000000  % originally 2000000 in the embedded TeX distribution in Therion for Windows

 

I also updated this in the Windows installer: https://github.com/therion/therion/actions/runs/25989238617

 

Martin

 

 

On Sun, Apr 19, 2026 at 7:20 AM <bruce@tomo.co.nz> wrote:

I’m exploring what might be causing this problem that I raised 3 weeks ago (copied below).

 

It has been affecting a project where we are migrating a lot of centrelines from On-Station to Therion.

Back in March this problem only affected exported pdf elevations that were over about 30 km in length.

Now we are closer to 63 km, it is also affecting exported pdf plans.

 

If I drill down to subsets of the project, there is no particular atomic subset that errors out.

Seems like when the subset drops below around 30 km the problem resolves itself and everything works OK.

 

Paring the layout used for exports down to specify only the scale of the output and then gradually adding statements, I have narrowed the culprit down to the colour map-fg statement.

 

Specifically, colour map-fg [97 86 38] is what seems to trigger the error.

Similarly colour map-fg scrap and colour map-fg map.

 

However colour map-fg 50 does not exhibit the problem for the 30-ish km centrelines.  An example below.

Neither does colour map-fg altitude (a colour legend is placed in the header, but the LRUD is not coloured – don’t know if that is expected or not).

 

This breaks down somewhat at 60-ish km and all pdf exports fail even with only a scale specified.

I have not figured out if this affects all my large projects, as I got interrupted by the map -survey option problem reported earlier.

 

Stepping back in time, based on limited testing, all versions back to at least therion 6.2.1 (2024-03-20) behave similarly).

 

Bruce

 

_______________________________________________
Therion mailing list
Therion@speleo.sk
https://mailman.speleo.sk/listinfo/therion