On 4/1/19 2:25 PM, Joel Kulesza wrote: > On Mon, Apr 1, 2019 at 10:22 AM Richard Kimberly Heck > <rikih...@lyx.org <mailto:rikih...@lyx.org>> wrote: > > On 3/31/19 8:13 PM, Uwe Stöhr wrote: > > Am 26.03.2019 um 20:58 schrieb Richard Kimberly Heck: > > > >> Just to be super clear: The compilers used are different, so the > >> binaries are different. > > > > How do youu build? I use MSVC 2015. > > I use gcc in a cross-platform environment. I.e., I compile for Windows > on Linux. > > > To be sure: is the process you (Riki) are describing the one on Line > 156 of Install.Win32? If not, can you please guide me to a > description of the process you're following (either your own > documentation in the repo or a more generic guidance document)?
No, my process is a bit different. I guess I should document it somewhere (besides here). But basically, you install mingw and all the same dependencies (though the mingw versions) you'd need for compiling LyX normally. (So, e.g., mingw32-qt5*, to be safe; mingw32-zlib; etc. I'm using the 32 bit versions now, but should try the 64 bit later.) Then it's: # mkdir build-mingw # cd build-mingw # bash ../development/cmake/scripts/cmingw That's a script that runs mingw's version of cmake with certain settings, which you can find in the script. Then, once that completes: # make # make package The latter builds a zip file with all necessary libraries included. Unfortunately, as it is now, it will only work on Fedora, because the paths seem to be different on different systems. But it could presumably be adapted to other systems. (The old version was for Ubuntu.) My own practice is then to copy the zip file over to Windows running in a VM and build the installer there. There's a version of NSIS that runs on Linux, but the installer uses Windows style paths, so it won't work on Linux without heavy modification, at least, and I'm not sure it will work at all. Why compile on Linux? It's a LOT easier. Riki