Il Mar 8 Feb 2022, 20:54 Michal Vlasák <lahc...@gmail.com> ha scritto:
> On Tue Feb 8, 2022 at 8:05 AM CET, luigi scarso wrote: > > Can you post a patch using the current luatex experimental repo at > > https://serveur-svn.lri.fr/svn/modhel/luatex/branches/experimental/ > > (see http://www.luatex.org/download.html ) ? > > Attached. > > I don't know how basing one patch on top of another works with TLpatches > and its .original files. I don't even even know if its manual or not, I > constructed the TLpatch file manually. > > >From the version I shared last time there are three changes: > > - byte swapping when loaded byte code has different byte order > - for simplicity I am wasting 8 bytes in each bytecode for detecting > byte order and actually verify that integers dump correctly > - I changed the "version" byte from 0 to 'T' so that it is explicit > that other Lua 5.3.6 bytecodes are incompatible with LuaTeX > > Neither the variable length encoding, neither byte swapping are compile > time options. I don't really want to get into preprocessor macros. This > seemed the simplest thing to work and shouldn't have much runtime cost > for most runs. > > Just for record, I am not a proponent of these changes, it's a > demonstration that it can and has to be done on the Lua level and that > it seemingly works. > > Also, LuaJIT bytecode is portable across all its architectures AFAIK, so > no problem with LuaJITTeX. > > Sorry, I also didn't test the LuaTeX repository build, it's hard for me > to cross compile and run binaries that excersise byte order issues > and/or 16/32/64 bit architectures problems. > > Michal > Ok , thank you very much. Maybe we can use qemu to check the 32/64 issue ? -- luigi