| Newsgroups |
gmane.comp.tex.live |
| Message-ID |
<[email protected]> |
Hello,
As an avid early adopter of pretest, I noticed a significant regression in pdflatex, both in performance and size of the generated PDF.
A minimal example:
\documentclass{article}
\usepackage{lipsum}
\begin{document}
\lipsum[1-2]
\end{document}
The output size difference scales with document length; if I replace “2” with “n”, I get the following behavior (size in bytes):
n | 2 | 5 | 10 | 20 | 50 | 100
2025 | 19519 | 21688 | 24721 | 29999 | 46467 | 72114
2026 | 20628 | 23951 | 29176 | 38459 | 67640 | 114502
I see a similar difference in scaling for the compilation time as reported by latexmk, with 2025 being nearly constant while 2026 doubles from 2 to 200.
2025:
> This is pdfTeX, Version 3.141592653-2.6-1.40.28 (TeX Live 2025) (preloaded format=pdflatex) \write18 enabled.
2026:
> This is pdfTeX, Version 3.141592653-2.6-1.40.29 (TeX Live 2026) (preloaded format=pdflatex) \write18 enabled.
This is on macOS 26.3 (arm64), in case this makes a difference. I do not see the same effect with lualatex, here the output size is identical between 2025 (1.22) and 2026 (1.24)
Let me know if you need more information (such as generated PDFs, which I don’t want to attach unasked-for).
Best,
Christian