[Tarantool-patches] [PATCH v3] lua/utils: fix fiber->fid print in cord_on_yield
Alexander V. Tikhonov
avtikhon at tarantool.org
Mon May 17 11:22:40 MSK 2021
Hi Cyrill, thanks for the patch, it LGTM.
On Fri, May 14, 2021 at 03:15:14PM +0300, Cyrill Gorcunov via Tarantool-patches wrote:
> On Fri, May 14, 2021 at 01:41:15PM +0300, Igor Munkin wrote:
> > Cyrill,
> >
> > Thanks for the fixes! Just a couple of nits.
> >
> > On 14.05.21, Cyrill Gorcunov via Tarantool-patches wrote:
> > > This fixes a nit in commit 6af473778
> > > (fiber: use uint64_t for fiber IDs).
> >
> > If you want to mention this commit, please use the full hash and enclose
>
> I don't understand why you need a full hash? When the reference is generated
> the git makes sure that id being generated is long enough among existing
> commits to be parseble. But sure, no problem.
>
> > the commit subject into single quotes: it's the common way used in repo.
> > You can find an example here[1] (but GitHub strips the hash while
> > rendering the webpage).
> >
> > >
> > > Since lua_pushfstring doesn't support %llu format it
> > > breaks the test
> >
> > Minor: It would be nice to refer Lua Reference Manual[2] then.
>
> Sure, thanks!
>
> > > + snprintf(buf, sizeof(buf),
> > > + "fiber %llu is switched while running the"
> > > + " compiled code (it's likely a function with"
> > > + " a yield underneath called via LuaJIT FFI)",
> > > + (long long)fiber()->fid);
> >
> > Why do you cast the value to long long (i.e. signed value), but use the
> > unsigned %llu modifier? This might be OK, but it's not clear to me.
>
> Because only _size_ of argument does matter, and language standart
> guarantees that signed and unsigned arguments are counterparts. Thus
> instead of big "unsigned long long" specificator you can write a
> shorthand "long long".
>
> Force pushed an update.
> ---
> From: Cyrill Gorcunov <gorcunov at gmail.com>
> Date: Fri, 14 May 2021 13:18:13 +0300
>
> This fixes a nit in commit 6af4737789042b1e6673c421686bb4394a61f4a1
> ("fiber: use uint64_t for fiber IDs").
>
> Since lua_pushfstring doesn't support %llu format [1] it
> breaks the test
>
> app-tap/gh-1700-abort-recording-on-fiber-switch.test.lua
>
> [1] https://www.lua.org/manual/5.1/manual.html#lua_pushfstring
>
> Signed-off-by: Cyrill Gorcunov <gorcunov at gmail.com>
> ---
> src/lua/utils.c | 20 +++++++++++++-------
> 1 file changed, 13 insertions(+), 7 deletions(-)
>
> diff --git a/src/lua/utils.c b/src/lua/utils.c
> index 0fbe700fc..3ce821374 100644
> --- a/src/lua/utils.c
> +++ b/src/lua/utils.c
> @@ -1341,6 +1341,7 @@ void cord_on_yield(void)
> * code misbehaviour and failures, so stop its execution.
> */
> if (unlikely(tvref(g->jit_base))) {
> + char buf[256];
> /*
> * XXX: mcode is executed only in scope of Lua
> * world and one can obtain the corresponding Lua
> @@ -1348,10 +1349,12 @@ void cord_on_yield(void)
> */
> struct lua_State *L = fiber()->storage.lua.stack;
> assert(L != NULL);
> - lua_pushfstring(L, "fiber %llu is switched while running the"
> - " compiled code (it's likely a function with"
> - " a yield underneath called via LuaJIT FFI)",
> - (long long)fiber()->fid);
> + snprintf(buf, sizeof(buf),
> + "fiber %llu is switched while running the"
> + " compiled code (it's likely a function with"
> + " a yield underneath called via LuaJIT FFI)",
> + (long long)fiber()->fid);
> + lua_pushstring(L, buf);
> if (g->panic)
> g->panic(L);
> exit(EXIT_FAILURE);
> @@ -1376,11 +1379,14 @@ void cord_on_yield(void)
> * GC hook is active and the platform is forced to stop.
> */
> if (unlikely(g->hookmask & HOOK_GC)) {
> + char buf[128];
> struct lua_State *L = fiber()->storage.lua.stack;
> assert(L != NULL);
> - lua_pushfstring(L, "fiber %d is switched while running GC"
> - " finalizer (i.e. __gc metamethod)",
> - fiber()->fid);
> + snprintf(buf, sizeof(buf),
> + "fiber %llu is switched while running GC"
> + " finalizer (i.e. __gc metamethod)",
> + (long long)fiber()->fid);
> + lua_pushstring(L, buf);
> if (g->panic)
> g->panic(L);
> exit(EXIT_FAILURE);
> --
> 2.31.1
>
More information about the Tarantool-patches
mailing list