[COMMIT] Set input FDs to non-blocking unconditionally, thank you Myrkraverk
7 years
Aidan Kehoe
Aidan Kehoe
Ar an deichiú lá de mí na Nollaig, scríobh Johann 'Myrkraverk' Oskarsson:
> On Sat, Dec 9, 2017 at 7:49 PM, Johann 'Myrkraverk' Oskarsson
> <johann(a)myrkraverk.com> wrote:
>
> > Ok, so I "worked around this problem" by setting the read end of the
> > pipe to nonblocking mode before calling connect_to_file_descriptor().
> >
>
> This seems to work fine for me. The pre-patched version is just one line,
> so I paste it here in its entirety.
I’ve gone ahead and applied that to 21.5. I have occasionally got hangs
especially when VM is using an external program for qp-decoding, and I suspect
your issue is at the root of them. I don’t see any downside to your change,
XEmacs code further up the call stack needs to be able to handle smaller reads
anyway. Thank you!
APPROVE COMMIT
NOTE: This patch has been committed
# HG changeset patch
# User Aidan Kehoe <kehoea(a)parhasard.net>
# Date 1513067231 0
# Tue Dec 12 08:27:11 2017 +0000
# Node ID affc195870d60c7f3f61d304cc01f205f63432d8
# Parent 0f7f0274157b50c71683368d0614b8912e2b70d7
Set input FDs to non-blocking unconditionally, thank you Myrkraverk
src/ChangeLog addition:
2017-12-12 Aidan Kehoe <kehoea(a)parhasard.net>
* process-unix.c (unix_init_process_io_handles):
Set input FDs to non-blocking unconditionally, see recent patch
posted to xemacs-beta from Johann "Myrkraverk" Oskarsson.
diff -r 0f7f0274157b -r affc195870d6 src/ChangeLog
--- a/src/ChangeLog Wed Dec 06 20:52:28 2017 +0000
+++ b/src/ChangeLog Tue Dec 12 08:27:11 2017 +0000
＠＠ -1,3 +1,9 ＠＠
+2017-12-12 Aidan Kehoe <kehoea(a)parhasard.net>
+
+ * process-unix.c (unix_init_process_io_handles):
+ Set input FDs to non-blocking unconditionally, see recent patch
+ posted to xemacs-beta from Johann "Myrkraverk" Oskarsson.
+
2017-12-02 Aidan Kehoe <kehoea(a)parhasard.net>
* bytecode.c (set_compiled_function_arglist):
diff -r 0f7f0274157b -r affc195870d6 src/process-unix.c
--- a/src/process-unix.c Wed Dec 06 20:52:28 2017 +0000
+++ b/src/process-unix.c Tue Dec 12 08:27:11 2017 +0000
＠＠ -831,6 +831,25 ＠＠
/* if sizeof(EMACS_INT) > sizeof(int) this truncates the value */
UNIX_DATA(p)->infd = (EMACS_INT) in;
UNIX_DATA(p)->errfd = (EMACS_INT) err;
+
+ /* In the event where the input file descriptor is a pipe, XEmacs as of
+ * 21.4.24 will hang on the read() system call when the pipe doesn't
+ * have enough to fill the incoming buffer.
+ *
+ * This will likely affect other kinds of blocking file descriptors also, so
+ * we unconditionally set the input fd to non-blocking mode. See:
+ * http://mid.xemacs.org/CAA_SOw__WYNiUO=4hHhiwg0LXq7++9Xr=tnBMgN5Uc8iAa+mvg...
+ * and the related thread.
+ */
+ {
+ int f = fcntl (UNIX_DATA(p)->infd, F_SETFL, O_NONBLOCK);
+ if (f == -1)
+ {
+ warn_when_safe (Qio_error, lisp_strerror (errno),
+ "failed setting pipe (fd %ld) to nonblocking mode",
+ (EMACS_INT) in);
+ }
+ }
}
/* Move the file descriptor FD so that its number is not less than MIN. *
--
‘As I sat looking up at the Guinness ad, I could never figure out /
How your man stayed up on the surfboard after forty pints of stout’
(C. Moore)
- 1 participants
- 0 comments
[COMMIT] Get X11 redisplay working again for the just-in-time Unicode characters.
7 years
Aidan Kehoe
Aidan Kehoe
APPROVE COMMIT
NOTE: This patch has been committed.
# HG changeset patch
# User Aidan Kehoe <kehoea(a)parhasard.net>
# Date 1512593548 0
# Wed Dec 06 20:52:28 2017 +0000
# Node ID 0f7f0274157b50c71683368d0614b8912e2b70d7
# Parent 1f54a2879725779c7da8c8a3b22de821a595e049
Get X11 redisplay working again for the just-in-time Unicode characters.
src/ChangeLog addition:
2017-12-06 Aidan Kehoe <kehoea(a)parhasard.net>
* unicode.c (allocate_jit_ucs_charset):
Ben removed the ccl-encode-to-ucs-2 CCL program even from the
non-unicode-internal build, and this meant that redisplay of the
just-in-time characters failed on X11.
Address this and simplify the code by leaving charset-registries
as nil for the just in time charsets, meaning the second-stage
code always gets invoked, and we don't need the CCL program.
diff -r 1f54a2879725 -r 0f7f0274157b src/ChangeLog
--- a/src/ChangeLog Sat Dec 02 08:28:51 2017 +0000
+++ b/src/ChangeLog Wed Dec 06 20:52:28 2017 +0000
＠＠ -13,6 +13,16 ＠＠
If KEY is a cons, loop using EXTERNAL_LIST_LOOP_3(), giving us
type and circularity checking.
+2017-12-06 Aidan Kehoe <kehoea(a)parhasard.net>
+
+ * unicode.c (allocate_jit_ucs_charset):
+ Ben removed the ccl-encode-to-ucs-2 CCL program even from the
+ non-unicode-internal build, and this meant that redisplay of the
+ just-in-time characters failed on X11.
+ Address this and simplify the code by leaving charset-registries
+ as nil for the just in time charsets, meaning the second-stage
+ code always gets invoked, and we don't need the CCL program.
+
2017-11-30 Aidan Kehoe <kehoea(a)parhasard.net>
* keymap.c (describe_command):
diff -r 1f54a2879725 -r 0f7f0274157b src/unicode.c
--- a/src/unicode.c Sat Dec 02 08:28:51 2017 +0000
+++ b/src/unicode.c Wed Dec 06 20:52:28 2017 +0000
＠＠ -1301,17 +1301,17 ＠＠
sprintf (setname, "jit-ucs-charset-%d", number_of_jit_charsets);
- Vcurrent_jit_charset = Fmake_charset
- (intern (setname), Vcharset_descr,
- nconc2 (list6 (Qcolumns, Qone, Qchars,
- make_fixnum (96),
- Qdimension, make_fixnum (2)),
- list4 (Qregistries, Qunicode_registries,
- /* not allowed to set non-nil tags when not yet
- initialized, for bootstrapping reasons; these
- get set in mule-charset.el */
- Qtags, initialized ? list2 (intern ("jit"), Qinternal)
- : Qnil)));
+ Vcurrent_jit_charset
+ = Fmake_charset (intern (setname), Vcharset_descr,
+ listu (Qcolumns, Qone, Qchars, make_fixnum (96),
+ Qdimension, make_fixnum (2),
+ /* not allowed to set non-nil tags when not yet
+ initialized, for bootstrapping reasons; these
+ get set in mule-charset.el */
+ Qtags, initialized ? list2 (intern ("jit"),
+ Qinternal) : Qnil,
+ Qunbound));
+
XCHARSET (Vcurrent_jit_charset)->jit_charset_p = 1;
last_allocated_jit_c1 = last_allocated_jit_c2 = 32;
--
‘As I sat looking up at the Guinness ad, I could never figure out /
How your man stayed up on the surfboard after forty pints of stout’
(C. Moore)
- 1 participants
- 0 comments
[COMMIT] Add some circularity checking, bytecode.c, eval.c, keymap.c
7 years
Aidan Kehoe
Aidan Kehoe
APPROVE COMMIT
NOTE: This patch has been committed
# HG changeset patch
# User Aidan Kehoe <kehoea(a)parhasard.net>
# Date 1512203331 0
# Sat Dec 02 08:28:51 2017 +0000
# Node ID 1f54a2879725779c7da8c8a3b22de821a595e049
# Parent 86ddcb2cb737db4e9518b1db0b7e4332a3d1b4d1
Add some circularity checking, bytecode.c, eval.c, keymap.c
src/ChangeLog addition:
2017-12-02 Aidan Kehoe <kehoea(a)parhasard.net>
* bytecode.c (set_compiled_function_arglist):
Check that NEW_ is a true list, avoid having a circular list in
the compiled funcction down the line.
* debug.c (Fset_debug_classes_to_check):
Use EXTERNAL_LIST_LOOP_3() looping through CLASSES for circularity
checking, leave the type checking to it.
* eval.c (Fmultiple_value_prog1):
ARGS is an external list, check for circularity and
well-formedness.
* keymap.c (Fsingle_key_description):
If KEY is a cons, loop using EXTERNAL_LIST_LOOP_3(), giving us
type and circularity checking.
tests/ChangeLog addition:
2017-12-02 Aidan Kehoe <kehoea(a)parhasard.net>
* automated/keymap-tests.el:
Test circularity checkcing in #'single-key-description.
diff -r 86ddcb2cb737 -r 1f54a2879725 src/ChangeLog
--- a/src/ChangeLog Fri Dec 01 19:46:24 2017 +0000
+++ b/src/ChangeLog Sat Dec 02 08:28:51 2017 +0000
＠＠ -1,3 +1,18 ＠＠
+2017-12-02 Aidan Kehoe <kehoea(a)parhasard.net>
+
+ * bytecode.c (set_compiled_function_arglist):
+ Check that NEW_ is a true list, avoid having a circular list in
+ the compiled funcction down the line.
+ * debug.c (Fset_debug_classes_to_check):
+ Use EXTERNAL_LIST_LOOP_3() looping through CLASSES for circularity
+ checking, leave the type checking to it.
+ * eval.c (Fmultiple_value_prog1):
+ ARGS is an external list, check for circularity and
+ well-formedness.
+ * keymap.c (Fsingle_key_description):
+ If KEY is a cons, loop using EXTERNAL_LIST_LOOP_3(), giving us
+ type and circularity checking.
+
2017-11-30 Aidan Kehoe <kehoea(a)parhasard.net>
* keymap.c (describe_command):
diff -r 86ddcb2cb737 -r 1f54a2879725 src/bytecode.c
--- a/src/bytecode.c Fri Dec 01 19:46:24 2017 +0000
+++ b/src/bytecode.c Sat Dec 02 08:28:51 2017 +0000
＠＠ -2681,7 +2681,7 ＠＠
static void
set_compiled_function_arglist (Lisp_Compiled_Function *f, Lisp_Object new_)
{
- CHECK_LIST (new_);
+ CHECK_TRUE_LIST (new_);
f->arglist = new_;
/* Recalculate the optimized version of the function, since this depends
diff -r 86ddcb2cb737 -r 1f54a2879725 src/debug.c
--- a/src/debug.c Fri Dec 01 19:46:24 2017 +0000
+++ b/src/debug.c Sat Dec 02 08:28:51 2017 +0000
＠＠ -135,16 +135,12 ＠＠
*/
(classes))
{
- Lisp_Object rest;
-
- CHECK_LIST (classes);
-
/* Make sure all objects in the list are valid. If anyone is not
valid, reject the entire list without doing anything. */
- LIST_LOOP (rest, classes)
+ EXTERNAL_LIST_LOOP_3 (elt, classes, rest)
{
- if (NILP (xemacs_debug_loop (X_VALIDATE, XCAR (rest), Qnil)))
- sferror ("Invalid object in class list", Qunbound);
+ if (NILP (xemacs_debug_loop (X_VALIDATE, elt, Qnil)))
+ sferror ("Invalid object in class list", elt);
}
LIST_LOOP (rest, classes)
diff -r 86ddcb2cb737 -r 1f54a2879725 src/eval.c
--- a/src/eval.c Fri Dec 01 19:46:24 2017 +0000
+++ b/src/eval.c Sat Dec 02 08:28:51 2017 +0000
＠＠ -5081,7 +5081,7 ＠＠
GCPRO1 (val);
{
- LIST_LOOP_2 (form, XCDR (args))
+ EXTERNAL_LIST_LOOP_2 (form, XCDR (args))
Feval (form);
}
diff -r 86ddcb2cb737 -r 1f54a2879725 src/keymap.c
--- a/src/keymap.c Fri Dec 01 19:46:24 2017 +0000
+++ b/src/keymap.c Sat Dec 02 08:28:51 2017 +0000
＠＠ -3593,10 +3593,8 ＠＠
{
DECLARE_EISTRING (bufp);
- Lisp_Object rest;
- LIST_LOOP (rest, key)
+ EXTERNAL_LIST_LOOP_3 (keysym, key, rest)
{
- Lisp_Object keysym = XCAR (rest);
if (EQ (keysym, Qcontrol)) eicat_ascii (bufp, "C-");
else if (EQ (keysym, Qctrl)) eicat_ascii (bufp, "C-");
else if (EQ (keysym, Qmeta)) eicat_ascii (bufp, "M-");
diff -r 86ddcb2cb737 -r 1f54a2879725 tests/ChangeLog
--- a/tests/ChangeLog Fri Dec 01 19:46:24 2017 +0000
+++ b/tests/ChangeLog Sat Dec 02 08:28:51 2017 +0000
＠＠ -1,3 +1,8 ＠＠
+2017-12-02 Aidan Kehoe <kehoea(a)parhasard.net>
+
+ * automated/keymap-tests.el:
+ Test circularity checkcing in #'single-key-description.
+
2017-11-23 Aidan Kehoe <kehoea(a)parhasard.net>
* automated/format-tests.el (args-out-of-range):
diff -r 86ddcb2cb737 -r 1f54a2879725 tests/automated/keymap-tests.el
--- a/tests/automated/keymap-tests.el Fri Dec 01 19:46:24 2017 +0000
+++ b/tests/automated/keymap-tests.el Sat Dec 02 08:28:51 2017 +0000
＠＠ -133,4 +133,10 ＠＠
pi))
(Check-Error wrong-type-argument (command-remapping 'describe-function-at-point
nil pi)))
+
+;; Not quite a keymap test, but this function is in keymap.c, so add a test
+;; here:
+(Check-Error circular-list (single-key-description '#1=(control shift . #1#)))
+
+;;; end of keymap-tests.el
--
‘As I sat looking up at the Guinness ad, I could never figure out /
How your man stayed up on the surfboard after forty pints of stout’
(C. Moore)
- 1 participants
- 0 comments
[COMMIT] Avoid a distracting warning about custom groups, about.el
7 years
Aidan Kehoe
Aidan Kehoe
APPROVE COMMIT
NOTE: This patch has been committed
# HG changeset patch
# User Aidan Kehoe <kehoea(a)parhasard.net>
# Date 1512157584 0
# Fri Dec 01 19:46:24 2017 +0000
# Node ID 86ddcb2cb737db4e9518b1db0b7e4332a3d1b4d1
# Parent dabe2da63f45bf454afdd10dd4e9b475a2ac8f06
Avoid a distracting warning about custom groups, about.el
lisp/ChangeLog addition:
2017-12-01 Aidan Kehoe <kehoea(a)parhasard.net>
* about.el (about): New custom group for this file.
* about.el (about-headline-face):
* about.el (about-link-face):
Provide and use a custom group for these faces, avoiding a
distracting warning when viewing the about page.
diff -r dabe2da63f45 -r 86ddcb2cb737 lisp/ChangeLog
--- a/lisp/ChangeLog Thu Nov 30 05:15:02 2017 +0000
+++ b/lisp/ChangeLog Fri Dec 01 19:46:24 2017 +0000
＠＠ -1,3 +1,11 ＠＠
+2017-12-01 Aidan Kehoe <kehoea(a)parhasard.net>
+
+ * about.el (about): New custom group for this file.
+ * about.el (about-headline-face):
+ * about.el (about-link-face):
+ Provide and use a custom group for these faces, avoiding a
+ distracting warning when viewing the about page.
+
2017-11-16 Aidan Kehoe <kehoea(a)parhasard.net>
* x-faces.el (x-library-search-path):
diff -r dabe2da63f45 -r 86ddcb2cb737 lisp/about.el
--- a/lisp/about.el Thu Nov 30 05:15:02 2017 +0000
+++ b/lisp/about.el Fri Dec 01 19:46:24 2017 +0000
＠＠ -49,13 +49,9 ＠＠
(require 'wid-edit)
-;; People in this list have their individual links from the main page,
-;; or from the `Legion' page. If they have an image, it should be
-;; named after the CAR of the list element (baw -> baw.png).
-;;
-;; If you add to this list, you'll want to update
-;; `about-personal-info' and `about-hackers', and add the name to one
-;; of the three mutually exclusive lists just below.
+;; It's unlikely anyone will customise this, but the warning about a
+;; non-existent group for the faces is distracting.
+(defgroup about nil "Customization group for the `About the Authors' page.")
(defface about-headline-face
'((((class color) (background dark))
＠＠ -71,7 +67,8 ＠＠
(((class grayscale) (background dark))
(:foreground "DimGray" :bold t))
(t (:bold t)))
- "Face used for color-highlighted headlines in the About page.")
+ "Face used for color-highlighted headlines in the About page."
+ :group 'about)
(defface about-link-face
'((((class color) (background dark))
＠＠ -86,7 +83,16 ＠＠
(((class grayscale) (background dark))
(:foreground "LightGray" :bold t :italic t :underline t))
(t (:underline t)))
- "Face used for links in the About page.")
+ "Face used for links in the About page."
+ :group 'about)
+
+;; People in this list have their individual links from the main page,
+;; or from the `Legion' page. If they have an image, it should be
+;; named after the CAR of the list element (baw -> baw.png).
+;;
+;; If you add to this list, you'll want to update
+;; `about-personal-info' and `about-hackers', and add the name to one
+;; of the three mutually exclusive lists just below.
(defvar xemacs-hackers
'(
--
‘As I sat looking up at the Guinness ad, I could never figure out /
How your man stayed up on the surfboard after forty pints of stout’
(C. Moore)
- 1 participants
- 0 comments
[AC 21.5] Shut up some clang warnings
7 years
Stephen J. Turnbull
Stephen J. Turnbull
APPROVE COMMIT 21.5
# HG changeset patch
# Parent a3aa49cd5d7559d6947d815a6aa78413e0010c30
Quiet clang about vdb_fault_handler prototype.
* * *
Quiet compiler warnings.
* * *
Combine with quiet-compiler, please.
diff -r a3aa49cd5d75 lwlib/ChangeLog
--- a/lwlib/ChangeLog Thu Nov 23 18:48:06 2017 +0000
+++ b/lwlib/ChangeLog Sat Nov 25 17:48:54 2017 +0900
＠＠ -1,3 +1,7 ＠＠
+2017-11-23 Stephen J. Turnbull <stephen(a)xemacs.org>
+
+ * lwlib-colors.c (xft_convert_color): Remove unused variable.
+
2017-11-12 Aidan Kehoe <kehoea(a)parhasard.net>
* lwlib-internal.h:
diff -r a3aa49cd5d75 lwlib/lwlib-colors.c
--- a/lwlib/lwlib-colors.c Thu Nov 23 18:48:06 2017 +0000
+++ b/lwlib/lwlib-colors.c Sat Nov 25 17:48:54 2017 +0900
＠＠ -355,7 +355,6 ＠＠
xft_convert_color (Display *dpy, Colormap cmap, Visual *visual, int c, int dim)
{
static XColor color; /* #### why is this static ?? */
- XftColor result;
color.pixel = c;
XQueryColor(dpy, cmap, &color);
diff -r a3aa49cd5d75 src/ChangeLog
--- a/src/ChangeLog Thu Nov 23 18:48:06 2017 +0000
+++ b/src/ChangeLog Sat Nov 25 17:48:54 2017 +0900
＠＠ -1,3 +1,19 ＠＠
+2017-11-24 Stephen J. Turnbull <stephen(a)xemacs.org>
+
+2017-11-23 Stephen J. Turnbull <stephen(a)xemacs.org>
+
+ * sysdep.c (total_data_usage):
+ Document that sbrk(0) usage is still OK.
+
+ * window.c (window_divider_width):
+ * window-impl.h (MODELINE_OFF_SHADOW_THICKNESS_ADJUSTED):
+ Use EMACS_INT_ABS to shut up compiler.
+
+ * lread.c (read1): Comment too-clever logic.
+
+ * vdb-posix.c (vdb_fault_handler):
+ Add prototype to quiet clang (at least 4.0 and 5.0).
+
2017-11-23 Aidan Kehoe <kehoea(a)parhasard.net>
* config.h.in:
diff -r a3aa49cd5d75 src/lread.c
--- a/src/lread.c Thu Nov 23 18:48:06 2017 +0000
+++ b/src/lread.c Sat Nov 25 17:48:54 2017 +0900
＠＠ -2644,6 +2644,9 ＠＠
while ((c = readchar (readcharfun)) >= 0
&& c >= '0' && c <= '9')
{
+ /* #### I think this logic is wrong. Skip can be in the
+ range 0x40000000-0xffffffff, the overflow will not
+ trigger, but this won't fit in a fixnum. */
skip = (10 * skip) + (c - '0');
if (oskip <= skip)
{
diff -r a3aa49cd5d75 src/sysdep.c
--- a/src/sysdep.c Thu Nov 23 18:48:06 2017 +0000
+++ b/src/sysdep.c Sat Nov 25 17:48:54 2017 +0900
＠＠ -1991,6 +1991,9 ＠＠
#endif
#ifndef WIN32_ANY
+ /* clang emits a deprecation warning but this usage documented OK.
+ On Linux, apparently can use /proc filesystem; on Mac OS X need to use
+ mach_vm_region_info which looks annoying. */
void *data_end = sbrk (0);
#else
void *data_end = maximum_address_seen;
diff -r a3aa49cd5d75 src/vdb-posix.c
--- a/src/vdb-posix.c Thu Nov 23 18:48:06 2017 +0000
+++ b/src/vdb-posix.c Sat Nov 25 17:48:54 2017 +0900
＠＠ -55,6 +55,7 ＠＠
sighandler_t segv_oact, bus_oact;
#endif /* USE_SIGNAL */
+void vdb_fault_handler (FAULT_HANDLER_ARGUMENTS);
void vdb_remove_signal_handler (void);
void
diff -r a3aa49cd5d75 src/window-impl.h
--- a/src/window-impl.h Thu Nov 23 18:48:06 2017 +0000
+++ b/src/window-impl.h Sat Nov 25 17:48:54 2017 +0900
＠＠ -269,12 +269,12 ＠＠
#define WINDOW_HAS_MODELINE_P(w) (!NILP (w->has_modeline_p))
#define MODELINE_OFF_SHADOW_THICKNESS_ADJUSTED(win) \
- abs ((!WINDOW_HAS_MODELINE_P (win) \
- ? ((XFIXNUM (win->modeline_shadow_thickness) > 1) \
- ? XFIXNUM (win->modeline_shadow_thickness) - 1 \
+ EMACS_INT_ABS ((!WINDOW_HAS_MODELINE_P (win) \
+ ? ((XFIXNUM (win->modeline_shadow_thickness) > 1) \
+ ? XFIXNUM (win->modeline_shadow_thickness) - 1 \
: ((XFIXNUM (win->modeline_shadow_thickness) < -1) \
? XFIXNUM (win->modeline_shadow_thickness) + 1 \
- : XFIXNUM (win->modeline_shadow_thickness))) \
+ : XFIXNUM (win->modeline_shadow_thickness))) \
: XFIXNUM (win->modeline_shadow_thickness)))
#define MODELINE_SHADOW_THICKNESS(win) \
diff -r a3aa49cd5d75 src/window.c
--- a/src/window.c Thu Nov 23 18:48:06 2017 +0000
+++ b/src/window.c Sat Nov 25 17:48:54 2017 +0900
＠＠ -1062,7 +1062,7 ＠＠
return
XFIXNUM (w->vertical_divider_line_width)
+ 2 * XFIXNUM (w->vertical_divider_spacing)
- + 2 * abs (XFIXNUM (w->vertical_divider_shadow_thickness));
+ + 2 * EMACS_INT_ABS (XFIXNUM (w->vertical_divider_shadow_thickness));
else
return XFIXNUM (w->vertical_divider_line_width) == 0 ? 0 : 1;
}
- 3 participants
- 5 comments