Package: emacs;
Reported by: Werner LEMBERG <wl <at> gnu.org>
Date: Sat, 29 Jun 2019 11:19:01 UTC
Severity: normal
Done: Stefan Monnier <monnier <at> iro.umontreal.ca>
Bug is archived. No further changes may be made.
View this message in rfc822 format
From: Stefan Monnier <monnier <at> iro.umontreal.ca> To: Werner LEMBERG <wl <at> gnu.org> Cc: eliz <at> gnu.org, 36431 <at> debbugs.gnu.org Subject: bug#36431: Crash in marker.c:337 Date: Tue, 02 Jul 2019 12:29:34 -0400
Stefan Monnier <monnier <at> iro.umontreal.ca> writes:
>> No, it doesn't fix the problem, unfortunately – mew doesn't start at
>> all with this patch; the error message shows a string that is
>> definitely cropped.
>
> Yes, I also saw other problems with the patch.
> I'll send an updated patch later.
Can you try the patch below?
Stefan
diff --git a/src/fileio.c b/src/fileio.c
index ed1d2aedf3..1fea93fa8e 100644
--- a/src/fileio.c
+++ b/src/fileio.c
@@ -3741,6 +3741,7 @@ by calling `format-decode', which see. */)
CHECK_CODING_SYSTEM (Vcoding_system_for_read);
Fset (Qbuffer_file_coding_system, Vcoding_system_for_read);
}
+ eassert (inserted == 0);
goto notfound;
}
@@ -3767,7 +3768,10 @@ by calling `format-decode', which see. */)
not_regular = 1;
if (! NILP (visit))
- goto notfound;
+ {
+ eassert (inserted == 0);
+ goto notfound;
+ }
if (! NILP (replace) || ! NILP (beg) || ! NILP (end))
xsignal2 (Qfile_error,
@@ -4435,19 +4439,6 @@ by calling `format-decode', which see. */)
if (how_much < 0)
report_file_error ("Read error", orig_filename);
- /* Make the text read part of the buffer. */
- GAP_SIZE -= inserted;
- GPT += inserted;
- GPT_BYTE += inserted;
- ZV += inserted;
- ZV_BYTE += inserted;
- Z += inserted;
- Z_BYTE += inserted;
-
- if (GAP_SIZE > 0)
- /* Put an anchor to ensure multi-byte form ends at gap. */
- *GPT_ADDR = 0;
-
notfound:
if (NILP (coding_system))
@@ -4457,6 +4448,7 @@ by calling `format-decode', which see. */)
Note that we can get here only if the buffer was empty
before the insertion. */
+ eassert (Z == BEG);
if (!NILP (Vcoding_system_for_read))
coding_system = Vcoding_system_for_read;
@@ -4477,6 +4469,10 @@ by calling `format-decode', which see. */)
bset_undo_list (current_buffer, Qt);
record_unwind_protect (decide_coding_unwind, unwind_data);
+ /* Make the text read part of the buffer. */
+ eassert (NILP (BVAR (current_buffer, enable_multibyte_characters)));
+ insert_from_gap_1 (inserted, inserted, false);
+
if (inserted > 0 && ! NILP (Vset_auto_coding_function))
{
coding_system = call2 (Vset_auto_coding_function,
@@ -4493,8 +4489,22 @@ by calling `format-decode', which see. */)
if (CONSP (coding_system))
coding_system = XCAR (coding_system);
}
- unbind_to (count1, Qnil);
+
+ /* Move the text back to the beginning of the gap.
+ Do it now, before we set the buffer back to multibyte, since the
+ bytes may very well not be valid for a multibyte buffer. */
+ set_point_both (BEG, BEG_BYTE);
+ /* In general this may have to move all the bytes, but here
+ this can't move more bytes than were moved during the execution
+ of Vset_auto_coding_function, which is normally 0 (because it
+ normally doesn't modify the buffer). */
+ move_gap_both (Z, Z_BYTE);
inserted = Z_BYTE - BEG_BYTE;
+ GAP_SIZE += inserted;
+ ZV = Z = GPT = BEG;
+ ZV_BYTE = Z_BYTE = GPT_BYTE = BEG_BYTE;
+
+ unbind_to (count1, Qnil);
}
if (NILP (coding_system))
@@ -4528,22 +4538,29 @@ by calling `format-decode', which see. */)
}
}
- coding.dst_multibyte = ! NILP (BVAR (current_buffer, enable_multibyte_characters));
+ eassert (PT == GPT);
+
+ coding.dst_multibyte
+ = ! NILP (BVAR (current_buffer, enable_multibyte_characters));
if (CODING_MAY_REQUIRE_DECODING (&coding)
&& (inserted > 0 || CODING_REQUIRE_FLUSHING (&coding)))
{
- move_gap_both (PT, PT_BYTE);
- GAP_SIZE += inserted;
- ZV_BYTE -= inserted;
- Z_BYTE -= inserted;
- ZV -= inserted;
- Z -= inserted;
+ /* Now we have all the new bytes at the beginning of the gap,
+ but `decode_coding_gap` needs them at the end of the gap, so
+ we need to move them.
+ FIXME: We should arrange for the bytes to be already at the right
+ place so we don't need to memmove them in the common case! */
+ memmove (GAP_END_ADDR - inserted, GPT_ADDR, inserted);
decode_coding_gap (&coding, inserted, inserted);
inserted = coding.produced_char;
coding_system = CODING_ID_NAME (coding.id);
}
else if (inserted > 0)
{
+ /* Make the text read part of the buffer. */
+ eassert (NILP (BVAR (current_buffer, enable_multibyte_characters)));
+ insert_from_gap_1 (inserted, inserted, false);
+
invalidate_buffer_caches (current_buffer, PT, PT + inserted);
adjust_after_insert (PT, PT_BYTE, PT + inserted, PT_BYTE + inserted,
inserted);
diff --git a/src/insdel.c b/src/insdel.c
index 85fffd8fd1..51371ee13c 100644
--- a/src/insdel.c
+++ b/src/insdel.c
@@ -115,7 +115,7 @@ gap_left (ptrdiff_t charpos, ptrdiff_t bytepos, bool newgap)
i = GPT_BYTE;
to = GAP_END_ADDR;
from = GPT_ADDR;
- new_s1 = GPT_BYTE;
+ new_s1 = GPT_BYTE; /* May point in the middle of multibyte sequences. */
/* Now copy the characters. To move the gap down,
copy characters up. */
@@ -133,6 +133,7 @@ gap_left (ptrdiff_t charpos, ptrdiff_t bytepos, bool newgap)
make_gap_smaller set inhibit-quit. */
if (QUITP)
{
+ /* FIXME: This can point in the middle of a multibyte character. */
bytepos = new_s1;
charpos = BYTE_TO_CHAR (bytepos);
break;
@@ -164,7 +165,7 @@ gap_right (ptrdiff_t charpos, ptrdiff_t bytepos)
{
register unsigned char *to, *from;
register ptrdiff_t i;
- ptrdiff_t new_s1;
+ ptrdiff_t new_s1; /* May point in the middle of multibyte sequences. */
BUF_COMPUTE_UNCHANGED (current_buffer, charpos, GPT);
@@ -189,6 +190,7 @@ gap_right (ptrdiff_t charpos, ptrdiff_t bytepos)
make_gap_smaller set inhibit-quit. */
if (QUITP)
{
+ /* FIXME: This can point in the middle of a multibyte character. */
bytepos = new_s1;
charpos = BYTE_TO_CHAR (bytepos);
break;
@@ -1072,6 +1074,31 @@ insert_from_string_1 (Lisp_Object string, ptrdiff_t pos, ptrdiff_t pos_byte,
/* Insert a sequence of NCHARS chars which occupy NBYTES bytes
starting at GAP_END_ADDR - NBYTES (if text_at_gap_tail) and at
+ GPT_ADDR (if not text_at_gap_tail).
+ Contrary to insert_from_gap, this does not invalidate any cache,
+ nor update any markers, nor record any buffer modification information
+ of any sort. */
+void
+insert_from_gap_1 (ptrdiff_t nchars, ptrdiff_t nbytes, bool text_at_gap_tail)
+{
+ GAP_SIZE -= nbytes;
+ if (! text_at_gap_tail)
+ {
+ GPT += nchars;
+ GPT_BYTE += nbytes;
+ }
+ ZV += nchars;
+ Z += nchars;
+ ZV_BYTE += nbytes;
+ Z_BYTE += nbytes;
+
+ /* Put an anchor to ensure multi-byte form ends at gap. */
+ if (GAP_SIZE > 0) *(GPT_ADDR) = 0;
+ eassert (GPT <= GPT_BYTE);
+}
+
+/* Insert a sequence of NCHARS chars which occupy NBYTES bytes
+ starting at GAP_END_ADDR - NBYTES (if text_at_gap_tail) and at
GPT_ADDR (if not text_at_gap_tail). */
void
@@ -1090,19 +1117,7 @@ insert_from_gap (ptrdiff_t nchars, ptrdiff_t nbytes, bool text_at_gap_tail)
record_insert (GPT, nchars);
modiff_incr (&MODIFF);
- GAP_SIZE -= nbytes;
- if (! text_at_gap_tail)
- {
- GPT += nchars;
- GPT_BYTE += nbytes;
- }
- ZV += nchars;
- Z += nchars;
- ZV_BYTE += nbytes;
- Z_BYTE += nbytes;
- if (GAP_SIZE > 0) *(GPT_ADDR) = 0; /* Put an anchor. */
-
- eassert (GPT <= GPT_BYTE);
+ insert_from_gap_1 (nchars, nbytes, text_at_gap_tail);
adjust_overlays_for_insert (ins_charpos, nchars);
adjust_markers_for_insert (ins_charpos, ins_bytepos,
diff --git a/src/lisp.h b/src/lisp.h
index a0619e64f2..1a1d8ee7e4 100644
--- a/src/lisp.h
+++ b/src/lisp.h
@@ -3667,6 +3667,7 @@ extern void insert (const char *, ptrdiff_t);
extern void insert_and_inherit (const char *, ptrdiff_t);
extern void insert_1_both (const char *, ptrdiff_t, ptrdiff_t,
bool, bool, bool);
+extern void insert_from_gap_1 (ptrdiff_t, ptrdiff_t, bool text_at_gap_tail);
extern void insert_from_gap (ptrdiff_t, ptrdiff_t, bool text_at_gap_tail);
extern void insert_from_string (Lisp_Object, ptrdiff_t, ptrdiff_t,
ptrdiff_t, ptrdiff_t, bool);
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.