JosephFerano/flan
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

The FFI case at -O0 too, where the allocas are still there

Browse Source 
Every other program in the table is run twice for the reason stated next to
them: at -O2 mem2reg launders a sloppy alloca, so -O0 is what tests the IR
actually emitted. The raylib case had been the exception, and it is the worst
one to exempt — five of its calls hand C the address of a local struct, which
is exactly the alloca that comment is about.

It passes as it stands. That is the point: the case that only ever ran
optimised was a coincidence away from hiding something.
This commit is contained in:
Joseph Ferano 2026-09-11 17:59:00 +07:00
parent 01603843c0
commit 16f1605580
1 changed files with 15 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
test/test_acceptance.ml
View File

@ -159,13 +159,21 @@ let () =
symmetric. What the last case cannot pin, because nothing raylib
computes without a GL context reads them, is width, height and mipmaps
against each other. *)
if Sys.command "ldconfig -p 2>/dev/null | grep -q libraylib" = 0 then
outputs "raylib ffi, headless" "programs/raylib-ffi.flan"
"17\n34\n51\n68\n\
6\n1\n4\n3\n\
7\n13\n17\n2\n4\n3.5\n7.25\n11.5\n13.75\n\
1\n1\n1\n1\n7\n\
7\n0\n17\n2\n4\n"
let raylib_out =
"17\n34\n51\n68\n\
6\n1\n4\n3\n\
7\n13\n17\n2\n4\n3.5\n7.25\n11.5\n13.75\n\
1\n1\n1\n1\n7\n\
7\n0\n17\n2\n4\n"
in
if Sys.command "ldconfig -p 2>/dev/null | grep -q libraylib" = 0 then begin
outputs "raylib ffi, headless" "programs/raylib-ffi.flan" raylib_out;
(* And at -O0, for the reason the rest of the table is: every struct
here crosses as (addr v) on a local, which is the alloca mem2reg
would launder before anyone noticed it was wrong. *)
outputs ~opt:"-O0" "raylib ffi, headless, -O0" "programs/raylib-ffi.flan"
raylib_out
end
else
print_endline "acceptance: skipping the raylib FFI case (no libraylib)";