Re: [PATCH] http.c: clear the 'finished' member once we are done with it

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi Junio,

On Mon, 23 May 2022, Junio C Hamano wrote:

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
>
> > I stumbled over the need for this while investigating the build failures
> > caused by upgrading Git for Windows' SDK's GCC to v12.x.
> >
> >> diff --git a/http.c b/http.c
> >> index 229da4d148..85437b1980 100644
> >> --- a/http.c
> >> +++ b/http.c
> >> @@ -1367,6 +1367,9 @@ void run_active_slot(struct active_request_slot *slot)
> >>  			select(max_fd+1, &readfds, &writefds, &excfds, &select_timeout);
> >>  		}
> >>  	}
> >> +
> >> +	if (slot->finished == &finished)
> >> +		slot->finished = NULL;
> >
> > First of all, I suspect that
> > https://github.com/git/git/blob/v2.36.1/http.c#L1207 makes sure that GCC's
> > complaint is not actually accurate: we always re-set `finished` to `NULL`
> > when getting an unused slot, so even if there is a left-over dangling
> > pointer, it is not actually used, ever.
> >
> > But we need something to pacify GCC. Let's look at your patch.
> >
> > The first thing to note is that this is not _quite_ thread-safe: between
>
> Does this part of the code ever run multi-threaded?

It calls into cURL, which I suspect has a multi-threaded mode of
operation, and we do use some callbacks in `http-walker.c` and I was
worried that these callbacks could be called via cURL calls. That's why I
am concerned about thread-safety. I have looked for a while and not found
any way that code in `http.c` or `http-walker.c` could set `finished` by
way of a cURL callback, but there is a lot of code there and I could have
missed something crucial.

> > If that analysis is correct, I would expect the correct solution to turn
> > `finished` into an attribute of the slot, and change its role to be a flag
> > that this slot is spoken for and cannot be re-used quite yet even if it is
> > not currently in use.
>
> I have a feeling that we've mentioned that at least twice (perhaps
> three times) in the recent past that it is in essense reverting what
> the "finished" change baa7b67d (HTTP slot reuse fixes, 2006-03-10)
> did.  We used to use the in-use bit of the slot as an indicator that
> the slot dispatched by run_active_slot() has finished (i.e. the
> in-use bit must be cleared when the request held in the struct is
> fully done), but that broke when a slot we are looking at in
> run_active_slot() is serviced (which makes in_use false), and then
> another request reuses the slot (now no longer in_use), before the
> control comes back to the loop.  "while (slot->in_use)" at the
> beginning of the loop was still true, but the original request the
> slot was being used for, the one that the run_active_slot() function
> cares about, has completed.
>
> So...

No, I suggested to replace the `finished` variable with an attribute (or
"field" or "member variable") of the slot, and to respect it when looking
for an unused slot, i.e. not only look for a slot whose `in_use` is 0 but
also require `reserved_for_use` to be 0. In essence, the
`run_active_slot()` function owns the slot, even if it is not marked as
`in_use`. That should address the same concern as baa7b67d but without
using a pointer to a local variable.

Ciao,
Dscho