From: Helmut Eller <eller.helmut@gmail.com>
To: Robert Pluim <rpluim@gmail.com>
Cc: 65211@debbugs.gnu.org
Subject: bug#65211: 30.0.50; threads and accept-process-output
Date: Fri, 11 Aug 2023 10:32:53 +0200 [thread overview]
Message-ID: <m2il9mq8qy.fsf@gmail.com> (raw)
In-Reply-To: <87sf8q6hzb.fsf@gmail.com> (Robert Pluim's message of "Thu, 10 Aug 2023 17:21:28 +0200")
[-- Attachment #1: Type: text/plain, Size: 1179 bytes --]
On Thu, Aug 10 2023, Robert Pluim wrote:
> Hmm. Maybe putting it in `deactivate_process' is better (not that Iʼve
> tested 😺)
After a night of sleep, it occurred to me that this could also be
considered a problem of incomplete initialization of the
fd_callback_info[fd] struct.
So I wondered who is supposed to initialize it. Unfortunately, there is
half a dozen of add_<foo>_fd functions that set various bits, some call
each other, some duplicate code. None of them sets the waiting_thread
slot. And there is no single point that clears everything out (except
init_process_emacs).
I also discovered that recompute_max_desc considers an fd unused if the
.flags field is zero. Luckily, recompute_max_desc is only called in
three places: delete_write_fd, delete_keyboard_wait_descriptor and
deactivate_process. The call in deactivate_process seems redundant as
it calls the other two anyway. It seems easier to re-initialize the
struct when it becomes unused than to initialize it when it is used the
first time.
So I'm proposing the patch below that moves the re-initialization to one
single place, namely: recompute_max_desc.
Helmut
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-Re-initialize-fd_callback_data-fd-more-thoroughly.patch --]
[-- Type: text/x-diff, Size: 3127 bytes --]
From 675a76b297e9f050f719489a1419f0278ad8e9ae Mon Sep 17 00:00:00 2001
From: Helmut Eller <eller.helmut@gmail.com>
Date: Fri, 11 Aug 2023 10:31:24 +0200
Subject: [PATCH] Re-initialize fd_callback_data[fd] more thoroughly
The waiting_thread field was not re-initialized properly when a a
closed file descriptor resued by another process. (bug#65211)
As recompute_max_desc must be called when a file descriptor becomes
unused, we can do some re-initialization there.
* src/process.c (recompute_max_desc): Take the fd that was just
deleted as argument fd. If the flags field is 0, reset the rest of
the struct too.
(delete_write_fd, delete_keyboard_wait_descriptor): Update callers
accordingly.
(delete_read_fd): This is now just an alias for
delete_keyboard_wait_descriptor.
(deactivate_process): Remove the call to recompute_max_desc, as
delete_read_fd is called anyway.
---
src/process.c | 46 ++++++++++++++++------------------------------
1 file changed, 16 insertions(+), 30 deletions(-)
diff --git a/src/process.c b/src/process.c
index 08cb810ec13..cd5fd97e9d6 100644
--- a/src/process.c
+++ b/src/process.c
@@ -507,13 +507,6 @@ void
delete_read_fd (int fd)
{
delete_keyboard_wait_descriptor (fd);
-
- eassert (0 <= fd && fd < FD_SETSIZE);
- if (fd_callback_info[fd].flags == 0)
- {
- fd_callback_info[fd].func = 0;
- fd_callback_info[fd].data = 0;
- }
}
/* Add a file descriptor FD to be monitored for when write is possible.
@@ -544,19 +537,22 @@ add_non_blocking_write_fd (int fd)
}
static void
-recompute_max_desc (void)
+recompute_max_desc (int fd)
{
- int fd;
-
+ eassert (0 <= fd && fd < FD_SETSIZE);
eassert (max_desc < FD_SETSIZE);
- for (fd = max_desc; fd >= 0; --fd)
- {
+
+ if (fd_callback_info[fd].flags == 0)
+ fd_callback_info[fd] = (struct fd_callback_data){ 0 };
+
+ if (fd == max_desc)
+ for (; fd >= 0; --fd)
if (fd_callback_info[fd].flags != 0)
- {
- max_desc = fd;
- break;
- }
- }
+ {
+ max_desc = fd;
+ break;
+ }
+
eassert (max_desc < FD_SETSIZE);
}
@@ -569,17 +565,10 @@ delete_write_fd (int fd)
if ((fd_callback_info[fd].flags & NON_BLOCKING_CONNECT_FD) != 0)
{
if (--num_pending_connects < 0)
- emacs_abort ();
+ emacs_abort ();
}
fd_callback_info[fd].flags &= ~(FOR_WRITE | NON_BLOCKING_CONNECT_FD);
- if (fd_callback_info[fd].flags == 0)
- {
- fd_callback_info[fd].func = 0;
- fd_callback_info[fd].data = 0;
-
- if (fd == max_desc)
- recompute_max_desc ();
- }
+ recompute_max_desc (fd);
}
static void
@@ -4809,8 +4798,6 @@ deactivate_process (Lisp_Object proc)
delete_read_fd (inchannel);
if ((fd_callback_info[inchannel].flags & NON_BLOCKING_CONNECT_FD) != 0)
delete_write_fd (inchannel);
- if (inchannel == max_desc)
- recompute_max_desc ();
}
}
@@ -8134,8 +8121,7 @@ delete_keyboard_wait_descriptor (int desc)
fd_callback_info[desc].flags &= ~(FOR_READ | KEYBOARD_FD | PROCESS_FD);
- if (desc == max_desc)
- recompute_max_desc ();
+ recompute_max_desc (desc);
#endif
}
--
2.39.2
prev parent reply other threads:[~2023-08-11 8:32 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-08-10 15:04 bug#65211: 30.0.50; threads and accept-process-output Helmut Eller
2023-08-10 15:21 ` Robert Pluim
2023-08-10 16:24 ` Helmut Eller
2023-08-11 8:32 ` Helmut Eller [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://www.gnu.org/software/emacs/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=m2il9mq8qy.fsf@gmail.com \
--to=eller.helmut@gmail.com \
--cc=65211@debbugs.gnu.org \
--cc=rpluim@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://git.savannah.gnu.org/cgit/emacs.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).