\endwrite token in newer version of PDFTeX - TeX

TopAnswers TeX

Meta

Databases

TeX

Code Golf

APL

C++

.net

db<>fiddle

Java

*nix

PHP

PowerShell

Python

Rust

टेक्-मराठी

Typst

Web Client Dev

Web Server Dev

\endwrite token in newer version of PDFTeX

add tag

user202729

In the newest version of PDFTeX, it looks like the behavior of the `\endwrite` token was changed somehow that the [method](https://tex.stackexchange.com/questions/609423/accessing-texs-internal-tokens) to access the token no longer work.

With some experiment:

```
\catcode`\  9
\endlinechar 32
\def \weird {    aaa \expandafter \weirda \expandafter { \iffalse } \fi }
\def \weirda #1  {  \expandafter \meaning  \noexpand }
\immediate\write-1{\weird}
bcdef
```
(run with `tex`, not `pdftex`, it segmentation fault in the latter as if `\meaning` is applied to unsafe `\endwrite` before the patch) -- looks like the meaning remains `\outer macro:`, but now `\noexpand` no longer "works" with `\endwrite` token -- normally `\meaning <safe \endwrite>` would result in `\relax`.

Question: is there any way to create the `\endwrite` token in TeX/PDFTeX outside of `\write` context?

I don't see any obvious way because it's not possible to duplicate the `\endwrite` token, and section 1372 of TeX source code:

![image.png](/image?hash=a634c33e3253b7768672d3d05d0088fabf4d42ccd548710c75e72c7fd8b14773)

tokens will be unconditionally gobbled until `\endwrite` itself is seen.

Top Answer

user202729

Answering my own question: I suspect the answer is no.

As mentioned in a comment, the patch being applied is exactly that `\noexpand\endwrite` will not result in the safe `\endwrite` anymore.

An `end write token` will only be inserted when `write` starts, and it will be removed later (the only way I can think of to interrupt the `get token` loop is to use invalid (cat 15) characters, and that doesn't allow breaking out of the loop either). With `\noexpand` gone there's very little things that can be done with `\endwrite` -- in particular it doesn't look possible to duplicate it, the error recovery on encountering a forbidden token while scanning argument seems to just terminate the argument scanning (around paragraph 392 but I haven't read carefully).


Anyway, looking through occurrences of `scanner status ← normal`, the only occurrences within expandable primitives I can find are `\ifx`, `\noexpand`, `\string`, `\meaning`. `\noexpand` is now useless and the other 3 just throw away the token.

I manage to segmentation fault `tex` again by doing `\ifx\endwrite\a` (not particularly useful however):

```tex
\outer\def\a{}
\def\weird{\expandafter\weirda\string}
\def\weirda#1{\ifx}
\immediate\write-1{\weird}\a
```

1 Answer