The change to the build flags solves the problem because hpcups and pxljr rely on being able to override jinit_color_converter. I think the fact that they're doing this points out a weakness in the libjpeg API - there should be some better way to register a callback to change the converter instead of hijacking the symbol from the library.
I think that for precise, turning off -Wl,-Bsymbolic-functions will be the right way to fix this. It's possible, but unlikely, that it will cause undesirable behavior regressions elsewhere... ultimately, it can only be a regression if some other package is *trying* to do this symbol interposition, has implemented it poorly, and users are being saved from that bug by symbolic-functions.
The change to the build flags solves the problem because hpcups and pxljr rely on being able to override jinit_color_ converter. I think the fact that they're doing this points out a weakness in the libjpeg API - there should be some better way to register a callback to change the converter instead of hijacking the symbol from the library.
I think that for precise, turning off -Wl,-Bsymbolic- functions will be the right way to fix this. It's possible, but unlikely, that it will cause undesirable behavior regressions elsewhere... ultimately, it can only be a regression if some other package is *trying* to do this symbol interposition, has implemented it poorly, and users are being saved from that bug by symbolic-functions.