[tex-k] segfault in texlive2020 pdftex/luatex/xetex/ptex/...

Karl Berry karl at freefriends.org
Tue Jun 2 00:31:37 CEST 2020

    ken> I can build tex so that it uses a bigger stack if that seems
    ken> desirable, by passing a suitable linker argument.

Ken, I think it would be desirable to build *tex with a sufficiently
large stack that the default expand_depth=10000 causes the TeX error
instead of sigsegv that DC saw. Ok?

We could also decrease the default, but since it only seems to be an
issue on Cygwin ... ?

    david> any usable relationship between the stack size as compiled in
    to the executable and the the expand_depth

Clearly there is a relationship, but I would not call it "usable", in
the sense of being predictable or quantifiable. And the relationship is
certainly platform- and compiler-specific in any case.

    Also is it documented anywhere? (I don't see in the web2c manual)

I'm guessing you mean expand_depth is not documented, not runtime stack
frame sizes :)?  I think you're right, it is not documented. I'll put in
some words about it, maybe in both web2c.texi and texmf.cnf.

    is it possible to default expand_depth to a value that
    makes the stack overflow less likely based on the compile time
    values used rather than experimentally setting it per machine. 

That's what the 10000 default value was the best guess at, and no one
has reported it before you, so I'm figuring that likely it's not
happening elsewhere. --thanks, karl.

More information about the tex-k mailing list.