Hi Thanks for the quick reply, I was hoping to discuss the randomseed on the PR, I've updated the patch on github
I'm hoping that reposts the patch here, if not I'll attach it Cheers Mike On Mon, 5 Jan 2026 at 10:36, Willy Tarreau <[email protected]> wrote: > Hi Lukas! > > On Mon, Jan 05, 2026 at 11:30:32AM +0100, Lukas Tribus wrote: > > Hello, > > > > On Fri, 2 Jan 2026 at 18:13, Willy Tarreau <[email protected]> wrote: > > > > > > Hi Mike, > > > > > > On Fri, Jan 02, 2026 at 03:23:03PM +0000, PR Bot wrote: > > > > From 53ceddc6a32ee7e3bde33b26e16a3fcbd44eb0df Mon Sep 17 00:00:00 > 2001 > > > > From: Mike Lothian <[email protected]> > > > > Date: Fri, 2 Jan 2026 14:38:10 +0000 > > > > Subject: [PATCH] MINOR: hlua: Add support for lua 5.5 > > > > > > > > --- > > > > src/hlua.c | 4 ++++ > > > > 1 file changed, 4 insertions(+) > > > > > > > > diff --git a/src/hlua.c b/src/hlua.c > > > > index 5109d3c04b33d..c3513cbca27ac 100644 > > > > --- a/src/hlua.c > > > > +++ b/src/hlua.c > > > > @@ -14027,7 +14027,11 @@ lua_State *hlua_init_state(int thread_num) > > > > struct prepend_path *pp; > > > > > > > > /* Init main lua stack. */ > > > > +#if defined(LUA_VERSION_NUM) && LUA_VERSION_NUM >= 505 > > > > + L = lua_newstate(hlua_alloc, &hlua_global_allocator, 0); > > > > +#else > > > > L = lua_newstate(hlua_alloc, &hlua_global_allocator); > > > > +#endif > > > > > > > > if (!L) { > > > > fprintf(stderr, > > > > > > Is it good enough to initialize lua_newstate with 0 seed? > > > > Before LUA 5.5, this function generated randomness internally. Since > > 5.5 it no longer does (requiring a seed), so by initializing it with a > > static 0 seed, we probably change some subsequent behavior that used > > to be random. > > Perhaps luaL_makeseed(0) or a haproxy provided randomness functions > > should be used instead of plain 0? > > I have no idea how this really affects LUA applications. > > > > This is the change in LUA: > > > https://github.com/lua/lua/commit/5a04f1851e0d42b4bcbb0af103490bc964e985aa > > Oh, great catch, you're right indeed! Maybe it would be better to > pass statistical_prng() in this case, as it looks like the previous > RNG was only time-related, hence not critical. > > Willy >
