Both binaries are excluded from the build in #2 because they do not compile on 0.15.2. These are real defects, not a toolchain artefact — CI on the targeted 0.15.2 reports them:
src/kg_cli.zig:43:26: error: root source file struct 'Io' has no member named 'getStdOut'
src/kg_server.zig:81:45: error: struct 'array_list.Aligned(kg_server.Triple,null)' has no member named 'init'
src/kg_server.zig:138:25: error: member function expected 2 argument(s), found 1
src/kg_server.zig:1151:43: error: expected type '*Io.Reader', found 'net.Server.Connection'
Three separate 0.14 → 0.15 changes:
std.io.getStdOut() / getStdIn() are gone. kg_cli.zig:43-44.
std.ArrayList is unmanaged. .init(allocator) no longer exists and append takes the allocator as its first argument. Call sites at lines 81, 229, 241, 469, 662 and every append/deinit/toOwnedSlice that follows them.
http.Server.init takes a reader, not a net.Server.Connection. kg_server.zig:1151.
That is a migration across roughly 1300 lines with many call sites, which is why it is not folded into #2 — the library is what other packages depend on and it builds and tests now; keeping the binaries in the default build would have kept the whole package unbuildable for the sake of two tools that have never run.
Worth noting for whoever picks it up: the reason none of this surfaced is that the repository had no build.zig and no workflow at all, so no compiler had ever read these files.
Both binaries are excluded from the build in #2 because they do not compile on 0.15.2. These are real defects, not a toolchain artefact — CI on the targeted 0.15.2 reports them:
Three separate 0.14 → 0.15 changes:
std.io.getStdOut()/getStdIn()are gone.kg_cli.zig:43-44.std.ArrayListis unmanaged..init(allocator)no longer exists andappendtakes the allocator as its first argument. Call sites at lines 81, 229, 241, 469, 662 and everyappend/deinit/toOwnedSlicethat follows them.http.Server.inittakes a reader, not anet.Server.Connection.kg_server.zig:1151.That is a migration across roughly 1300 lines with many call sites, which is why it is not folded into #2 — the library is what other packages depend on and it builds and tests now; keeping the binaries in the default build would have kept the whole package unbuildable for the sake of two tools that have never run.
Worth noting for whoever picks it up: the reason none of this surfaced is that the repository had no
build.zigand no workflow at all, so no compiler had ever read these files.