On 02/03/2015 08:36, Ulrike Fischer wrote:
> This doesn't work. Regardless where I put the \RequirePackage{etex},
> I get the same error.

That's odd, it worked for me. (I'll see if I can reproduce)
But in any case that isn't a viable fix it was just intended to be an
indicator of where the problem lies.

> The only thing that works is suppressing etex altogether with
>
> \expandafter\def\[log in to unmask]\endcsname{}
>
>
>>> Alternatively etex.sty could be adjusted to recognise the new
>>> format (for example an etex.sty that did nothing at all would
>>> work in this case)
> That would certainly a good idea.

Yes, probably something like this. Doing nothing at all works in this
case and in fact in the vast majority of cases that I can see in a
texlive 2014 tree, as etex.sty is just loaded so that \newcount doesn't
run out of registers at 255. However there are some packages that may
load etex and then expect the exact same implementation or a definition
of `\loccountblk` or something. It's hard to find any uses of
`\loccountblk` outside test files, but the internet is a big place...

So the original idea was to allow etex.sty to over-write the allocation
system to give exactly the old etex.sty behaviour. As far as I can see
that should work so long as etex.sty is loaded before extended registers
are allocated but I'll see if I can reproduce your failure, but in any
case it ought to be possible to make your original document work with no
change to the document. Details to be determined...

David