Forum Bugs

Crash in Mercury runtime (Prince 15.3)

Stefan Daschek
From time to time we see Prince 15.3 (running on Ubuntu 22.04) crashing with messages like:

*** Mercury runtime: caught segmentation violation ***
cause: address not mapped to object
PC at signal: 94327973681251 (55ca715d1863)
address involved: 0x3f8581e
This may have been caused by a stack overflow, due to unbounded recursion.
exiting from signal handler

The crash happens randomly (our job queue retries failed jobs more or less immediately, and we’ve not yet seen the same job crash twice).

So it could be the same issue as reported in https://www.princexml.com/forum/topic/4955/css-color-scheme-property-leads-to-mercury-runtime-caught ? Or has this alread been fixed in 15.3?

mikeday
It could be the same issue, can you try the latest builds?
Stefan Daschek
How stable are those builds in general, and how much do they differ from the latest release? We have to be rather cautious not to introduce incompatibilities in our system.
mikeday
Well if they (hopefully) don't crash then they are more stable. :D

The main difference is the initial support for CSS Grid Layout, so if you have documents that depend on grid layout not working and being treated identically to regular block layout then this will be a breaking change.
mikeday
We will issue a Prince 15.4 release shortly that also includes this fix.
mikeday
Prince 15.4 is now available, please let me know if it does not resolve the issue.
Stefan Daschek
Thanks! We updated our workers to use 15.4 – I’ll let you know if the issue remains.

(Rather unrelated: Our regression tests showed very slight differences regarding the rendering of embedded images in 15.4 – only a handful of pixels at the very edge of the image though, probably related to changes in anti-aliasing. I assume this is expected?)
mikeday
Prince 15.4 does include some minor bug fixes for SVG scaling; please email me (mikeday@yeslogic.com) a test document if you have concerns about the differences in output.