Treat ptex as eptex, uptex as euptex in plain

[aminophen]

Preparing https://github.com/texjporg/texjporg-testing, I noticed that the latex format does not like "e" (when "eptex" is specified, it starts building eplatex.fmt and fails)

checkengines    = checkengines
  or {"ptex", "uptex"}
checkformat  = "latex"

but the plain format needs explicit "e" (when "ptex" is specified, it crashes because the regression-test.tex always requires e-TeX)

checkengines    = checkengines
  or {"eptex", "euptex"}
checkformat  = "tex"

It would be nice if checkengines and stdengine for plain treat ptex as eptex and uptex as euptex.

[josephwright]

One could do testing without regression-test.tex, see for example the LaTeX2e test suite (uses it's own test2e.tex/test209.tex).

One should only need something like

specialformats = specialformats or  { }
specialformats.tex= specialformats.tex or {
    ptex  = {binary = "eptex"},
    uptex = {binary = "euptex"}
  }

which is much the same as we do for LaTeX: I could add to the default, but for plain testing I'm not so sure. (pLaTeX requires e-TeX so it's fine to force there ...).

In some ways I think one could argue this should be 'fixed at the engine end': pdfetex became just pdftex some years ago ...

[aminophen]

One could do testing without regression-test.tex

Oh, I didn't know that. Also, thanks for the information about specialformats. OK, I think this issue is invalid enough, closing.

one could argue this should be 'fixed at the engine end

Discussion already started some years ago (texjporg/tex-jp-build#32), but still need testing whether e-pTeX is completely (= bug-freely) upward compatible with pTeX. Also, which name (eptex vs ptex) should remain is still controversial ...

[josephwright]

We could perhaps make it clear that one can use a different set of macros provided they output the same markers in the log.

On the e-pTeX business: my main concern was that plain is different to LaTeX, and I don't want to set anything up that's misleading. What the Japanese community feel is right for e-pTeX vs pTeX is no an area I can really comment on!

[aminophen]

We could perhaps make it clear that one can use a different set of macros provided they output the same markers in the log.

That would be great! (then I think I can create some test on pTeX vs e-pTeX, which will make the testing much easier for us Japanese...)

[davidcarlisle]

I just hit something similar in the iftex bundle where most tests use latex or e-engines but I wanted one configuration with a few tests without e-tex so I could test \ifetex I ended up putting

ifx\numexpr\undefined
\def\START{\immediate\write-1{START-TEST-LOG}}
\let\zzend\end
\def\end{\immediate\write-1{END-TEST-LOG}\zzend}
\else
\input regression-test
\fi

at the top of the test files in https://github.com/latex3/iftex/tree/master/testfiles-plain

I did wonder if I should suggest that regression-test.tex defines some really basic setup, perhaps just \START \END and \TYPE if it detects etex isn't present.