|
From: Joel H. <jo...@ai...> - 2025-12-09 15:28:07
|
On 09/12/2025 00:19, Philipp Klaus Krause wrote: > Am 09.12.25 um 00:20 schrieb joel--- via sdcc-devel: >> Dear SDCC Developers, >> >> I'm happy to announce that I have successfully packaged SDCC v4.5.0 >> for Msys2's >> MINGW-packages: >> > > Thanks. This comes at the right time. We have not yet started the freeze > for 4.6.0 (there is not even a proposed schedule yet), so we still have > plenty of time to get your fixes in for the 4.6.0 release. > > I'll have a closer look at the details later. >> >> 1. Problems with "#define B" in ucsim. > > I had run into this when cross-compiling myself, buthadn't tracked it > down yet. I wonder if it would be sufficient to change the order of > #include, so standard headers (and thus GCC includes) come before the > #define B instead of changing B everywhere. I guess ultimately Daniel > will have to make this decision (and maybe we'll want to complain to > GCC, too). I think it would be a problem no matter which order the #includes occur. So long as "B" is #define'd somewhere, then "B" cannot be used anywhere without being replaced. >> 3. Problems with the assembler Makefile rules affecting parallel >> builds. >> On Msys2, single threaded build is unbearable slow. > > Hmm. I wonder why we don't run into this problem in the normal GNU/Linux > builds. I regularly compile with -j 18 on my machine, and in the farm we > do nightly compiles with -j 80 on two machines. This problem affects out-of-tree builds. There is a race condition between different threads copying and deleting the same source file into the build tree for different library variants being built in parallel. By the way, I forgot to mention that in-tree builds also have problems with installing even more spurious files. This can be seen in the file list of the Arch package: https://archlinux.org/packages/extra/x86_64/sdcc/ For example, all the .checkdevices and .deps directories. > I don't see patches with number 5 or higher at the link you provided. The rest of the issues were worked around in the PKGBUILD script: https://github.com/jhol/MINGW-packages/blob/d30aa8ae6d2004e7fbf79711ec800f074bf425d7/mingw-w64-sdcc/PKGBUILD Thanks Joel |