bug#15045: Point jumps inappropriately around time of Semantic lexing

[Barry OReilly <gundaetiapo@gmail.com>]

Eli:
> Can you identify the area in this trace where the unwarranted scroll
> was visible?

Barry:
> but redisplay didn't need to scroll

Undesired scrolling is a downstream symptom. Upstream from it is point
visibly moving around inappropriately. I have scroll-margin set to 4,
so when point briefly moves into the top scroll-margin, I get
undesired scrolling. I don't know that I've seen point move to outside
the visible part of the buffer. If so, then users with scroll-margin 0
won't see undesired scrolling, but would still see point moving
around.

Thus, I'm debugging the symptom of point moving.

> A, B, and C are local variables whose meaning depends on the email
> in which they're used.

Ah, so it's not in "thread local storage".

> in Fredisplay, walk the specpdl stack looking for a
> save_excursion_restore where the saved position is different from
> the current value of point in that buffer.

Thanks for the tip. I tested this change and it seems to implement
what you described.

diff --git a/src/dispnew.c b/src/dispnew.c
index 522a0e6..cf0103e 100644
--- a/src/dispnew.c
+++ b/src/dispnew.c
@@ -5815,6 +5815,32 @@ immediately by pending input.  */)
   (Lisp_Object force)
 {
   ptrdiff_t count;
+  bool noninteractive_old = noninteractive;
+  noninteractive = true;
+  Lisp_Object curPtMarker = Fpoint_marker();
+  union specbinding *pdl = specpdl_ptr;
+  while (pdl > specpdl)
+     {
+        --pdl;
+        if (pdl->kind == SPECPDL_UNWIND
+            && pdl->unwind.func == save_excursion_restore
+            && ! EQ (Fequal (XSAVE_OBJECT (pdl->unwind.arg, 0),
curPtMarker), Qt))
+           {
+              { struct timespec debug_ts; char debug_dateStr[20]; {
clock_gettime(CLOCK_REALTIME, &debug_ts); struct tm mytm;
localtime_r(&debug_ts.tv_sec, &mytm); strftime(debug_dateStr, 20,
"%Y-%m-%dT%H:%M:%S", &mytm)
+                 printf( "%s.%09ld|pid:%d|tid:%ld|%s|%d| DEBUG: Found
save_excursion_restore with mismatched point markers ", // TODO:
debugging
+                         debug_dateStr, debug_ts.tv_nsec, getpid(),
pthread_self(), __FILE__, __LINE__ ); }
+              Fprin1(XSAVE_OBJECT (pdl->unwind.arg, 0), Qnil);
+              Fprin1(curPtMarker, Qnil);
+              printf("\n");
+              Fbacktrace();
+              fflush(stdout);
+           }
+     }
+  /* { struct timespec debug_ts; char debug_dateStr[20]; {
clock_gettime(CLOCK_REALTIME, &debug_ts); struct tm mytm;
localtime_r(&debug_ts.tv_sec, &mytm); strftime(debug_dateStr, 20,
"%Y-%m-%dT%H:%M:%S", &mytm); } */
+  /*      printf( "%s.%09ld|pid:%d|tid:%ld|%s|%d| DEBUG: redisplay
\n", // TODO: debugging */
+  /*                        debug_dateStr, debug_ts.tv_nsec,
getpid(), pthread_self(), __FILE__, __LINE__ ); fflush(stdout); } */
+  /* Fbacktrace(); */
+  noninteractive = noninteractive_old;

   swallow_events (1);
   if ((detect_input_pending_run_timers (1)

We'll see what I get next time it comes up.