File 0157-Correct-printing-of-Unicode-atoms-in-crash... of Package erlang

Overview Repositories Revisions Requests Users Attributes Meta

File 0157-Correct-printing-of-Unicode-atoms-in-crash-dumps.patch of Package erlang (Revision 9e7ef33ae9bf36ae61b08a3391e2483a)

Currently displaying revision 9e7ef33ae9bf36ae61b08a3391e2483a , Show latest

From 5fc8d39e2c976f88d06c26de23422bee6f5ad2ab Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Mon, 31 May 2021 08:11:10 +0200
Subject: [PATCH] Correct printing of Unicode atoms in crash dumps

Atoms containing Unicode code points greater than 255 would not
be printed correctly in crash dumps.

This commit corrects the printing of Unicode atoms done by the
runtime-internal function `erts_printf()` (used when producing crash
dumps) and also updates `crashdump_viewer` to correctly handle such
atoms in function names in stack backtraces.
---
 erts/emulator/beam/erl_printf_term.c          | 95 +++++++++++++++++--
 lib/observer/src/crashdump_viewer.erl         |  3 +-
 lib/observer/src/observer_html_lib.erl        | 13 ++-
 .../test/crashdump_helper_unicode.erl         |  8 +-
 lib/observer/test/crashdump_viewer_SUITE.erl  | 19 +++-
 5 files changed, 123 insertions(+), 15 deletions(-)

diff --git a/erts/emulator/beam/erl_printf_term.c b/erts/emulator/beam/erl_printf_term.c
index 60bb09f0c6..47bbf588be 100644
--- a/erts/emulator/beam/erl_printf_term.c
+++ b/erts/emulator/beam/erl_printf_term.c
@@ -139,8 +139,6 @@ do {                                                        \
 #define IS_CNTRL(c)  ((c) < ' ' || (c) >= 127)
 #endif
 
-#define IS_PRINT(c)  (!IS_CNTRL(c))
-
 /* return 0 if list is not a non-empty flat list of printable characters */
 
 static int
@@ -185,7 +183,36 @@ static int is_printable_ascii(byte* bytep, Uint bytesize, Uint bitoffs)
     return 1;
 }
 
-/* print a atom doing what quoting is necessary */
+/*
+ * Helper function for print_atom_name(). Not generally useful.
+ */
+static ERTS_INLINE int latin1_char(int c1, int c2)
+{
+    if ((c1 & 0x80) == 0) {
+        /* Plain old 7-bit ASCII. */
+        return c1;
+    } else if ((c1 & 0xE0) == 0xC0) {
+        /* Unicode code points from 0x80 through 0x7FF. */
+        ASSERT((c2 & 0xC0) == 0x80);
+        return (c1 & 0x1F) << 6 | (c2 & 0x3F);
+    } else if ((c1 & 0xC0) == 0x80) {
+        /* A continutation byte in a utf8 sequence. Pretend that it is
+         * a character that is allowed in an atom. */
+        return 'a';
+    } else {
+        /* The start of a utf8 sequence comprising three or four
+         * bytes. Always needs quoting. */
+        return 0;
+    }
+}
+
+/*
+ * Print a atom, quoting it if necessary.
+ *
+ * Atoms are encoded in utf8. Since we have full control over creation
+ * of atoms, the utf8 encoding is always correct and there is no need
+ * to check for errors.
+ */
 static int print_atom_name(fmtfn_t fn, void* arg, Eterm atom, long *dcount)
 {
     int n, i;
@@ -195,6 +222,7 @@ static int print_atom_name(fmtfn_t fn, void* arg, Eterm atom, long *dcount)
     byte *s;
     byte *cpos;
     int c;
+    int lc;
 
     res = 0;
     i = atom_val(atom);
@@ -212,27 +240,47 @@ static int print_atom_name(fmtfn_t fn, void* arg, Eterm atom, long *dcount)
     *dcount -= atom_tab(i)->len;
 
     if (n == 0) {
+        /* The empty atom: '' */
 	PRINT_STRING(res, fn, arg, "''");
 	return res;
     }
 
+    /*
+     * Find out whether the atom will need quoting. Quoting is not necessary
+     * if the following applies:
+     *
+     *   - The first character is a lowercase letter in the Latin-1 code
+     *     block (0-255).
+     *
+     *   - All other characters are either alphanumeric characters in
+     *     the Latin-1 code block or the character '_'.
+     */
 
     need_quote = 0;
     cpos = s;
     pos = n - 1;
-
     c = *cpos++;
-    if (!IS_LOWER(c))
+    lc = latin1_char(c, *cpos);
+    if (!IS_LOWER(lc))
 	need_quote++;
     else {
 	while (pos--) {
 	    c = *cpos++;
-	    if (!IS_ALNUM(c) && (c != '_')) {
+            lc = latin1_char(c, *cpos);
+	    if (!IS_ALNUM(lc) && lc != '_') {
 		need_quote++;
 		break;
 	    }
 	}
     }
+
+    /*
+     * Now output the atom, including single quotes if needed.
+     *
+     * Control characters (including the range 128-159) must
+     * be specially printed. Therefore, we must do a partial
+     * decoding of the utf8 encoding.
+     */
     cpos = s;
     pos = n;
     if (need_quote)
@@ -249,12 +297,40 @@ static int print_atom_name(fmtfn_t fn, void* arg, Eterm atom, long *dcount)
 	case '\b': PRINT_STRING(res, fn, arg, "\\b"); break;
 	case '\v': PRINT_STRING(res, fn, arg, "\\v"); break;
 	default:
-	    if (IS_CNTRL(c)) {
+            if (c < ' ') {
+                /* ASCII control character (0-31). */
 		PRINT_CHAR(res, fn, arg, '\\');
 		PRINT_UWORD(res, fn, arg, 'o', 1, 3, (ErlPfUWord) c);
-	    }
-	    else
+            } else if (c >= 0x80) {
+                /* A multi-byte utf8-encoded code point. Determine the
+                 * length of the sequence. */
+                int n;
+                if ((c & 0xE0) == 0xC0) {
+                    n = 2;
+                } else if ((c & 0xF0) == 0xE0) {
+                    n = 3;
+                } else {
+                    ASSERT((c & 0xF8) == 0xF0);
+                    n = 4;
+                }
+                ASSERT(pos - n + 1 >= 0);
+
+                if (c == 0xC2 && *cpos < 0xA0) {
+                    /* Extended ASCII control character (128-159). */
+                    ASSERT(pos > 0);
+                    ASSERT(0x80 <= *cpos);
+                    PRINT_CHAR(res, fn, arg, '\\');
+                    PRINT_UWORD(res, fn, arg, 'o', 1, 3, (ErlPfUWord) *cpos);
+                    pos--, cpos++;
+                } else {
+                    PRINT_BUF(res, fn, arg, cpos-1, n);
+                    cpos += n - 1;
+                    pos -= n - 1;
+                }
+            } else {
+                /* Printable ASCII character. */
 		PRINT_CHAR(res, fn, arg, (char) c);
+            }
 	    break;
 	}
     }
@@ -263,7 +339,6 @@ static int print_atom_name(fmtfn_t fn, void* arg, Eterm atom, long *dcount)
     return res;
 }
 
-
 #define PRT_BAR                ((Eterm) 0)
 #define PRT_COMMA              ((Eterm) 1)
 #define PRT_CLOSE_LIST         ((Eterm) 2)
diff --git a/lib/observer/src/crashdump_viewer.erl b/lib/observer/src/crashdump_viewer.erl
index 2e3403325c..9ad281beeb 100644
--- a/lib/observer/src/crashdump_viewer.erl
+++ b/lib/observer/src/crashdump_viewer.erl
@@ -2943,7 +2943,8 @@ parse_term([$p|Line0], _, D) ->			%Port.
     {Port,Line} = get_id(Line0),
     {['#CDVPort'|Port],Line,D};
 parse_term([$S|Str0], _, D) ->			%Information string.
-    Str = lists:reverse(skip_blanks(lists:reverse(Str0))),
+    Str1 = byte_list_to_string(Str0),
+    Str = lists:reverse(skip_blanks(lists:reverse(Str1))),
     {Str,[],D};
 parse_term([$D|Line0], DecodeOpts, D) ->                 %DistExternal
     try
diff --git a/lib/observer/src/observer_html_lib.erl b/lib/observer/src/observer_html_lib.erl
index 4c92a8faab..4cce12dde8 100644
--- a/lib/observer/src/observer_html_lib.erl
+++ b/lib/observer/src/observer_html_lib.erl
@@ -138,8 +138,17 @@ msgq_table(Tab,Msg0, Id, Even, Colors) ->
     tr(color(Even, Colors),[td(integer_to_list(Id)), td(pre(Msg))]).
 
 stackdump_table(Tab,{Label0,Term0},Even, Colors) ->
-    Label = io_lib:format("~w",[Label0]),
-    Term = all_or_expand(Tab,Term0),
+    Label = io_lib:format("~ts",[Label0]),
+    Term = case atom_to_list(Label0) of
+               "y" ++ _ ->
+                   %% Any term is possible, including huge ones.
+                   all_or_expand(Tab,Term0);
+               _ ->
+                   %% Return address or catch tag. It is known to be a
+                   %% flat list, shortish, possibly containing characters
+                   %% greater than 255.
+                   href_proc_port(Term0)
+           end,
     tr(color(Even, Colors), [td("VALIGN=center",pre(Label)), td(pre(Term))]).
 
 dict_table(Tab,{Key0,Value0}, Even, Colors) ->
diff --git a/lib/observer/test/crashdump_helper_unicode.erl b/lib/observer/test/crashdump_helper_unicode.erl
index 60c3d20315..59cebef4fb 100644
--- a/lib/observer/test/crashdump_helper_unicode.erl
+++ b/lib/observer/test/crashdump_helper_unicode.erl
@@ -1,6 +1,6 @@
 -module(crashdump_helper_unicode).
 -behaviour(gen_server).
--export([start/0, init/1, handle_call/3, handle_cast/2]).
+-export([start/0, init/1, handle_call/3, handle_cast/2, 'спутник'/0]).
 -record(state, {s,a,b,lb}).
 
 start() ->
@@ -8,6 +8,9 @@ start() ->
 
 init([]) ->
     process_flag(trap_exit, true),
+    process_flag(save_calls, 10),
+    erlang:yield(),
+    ?MODULE:'спутник'(),
     ets:new('tab_αβ',[set,named_table]),
     Bin = <<"bin αβ"/utf8>>,
     LongBin = <<"long bin αβ - a utf8 binary which can be expanded αβ"/utf8>>,
@@ -20,3 +23,6 @@ handle_call(_Info, _From, State) ->
     {reply, ok, State}.
 handle_cast(_Info, State) ->
     {noreply, State}.
+
+'спутник'() ->
+    ok.
diff --git a/lib/observer/test/crashdump_viewer_SUITE.erl b/lib/observer/test/crashdump_viewer_SUITE.erl
index e4c4c4decb..7ed58a143c 100644
--- a/lib/observer/test/crashdump_viewer_SUITE.erl
+++ b/lib/observer/test/crashdump_viewer_SUITE.erl
@@ -593,9 +593,17 @@ special(File,Procs) ->
             #proc{pid=Pid0} =
                 lists:keyfind("'unicode_reg_name_αβ'",#proc.name,Procs),
             Pid = pid_to_list(Pid0),
-	    {ok,#proc{},[]} = crashdump_viewer:proc_details(Pid),
+	    {ok,Proc,[]} = crashdump_viewer:proc_details(Pid),
+            #proc{last_calls=LastCalls,stack_dump=Stk} = Proc,
             io:format("  unicode registered name ok",[]),
 
+            ["crashdump_helper_unicode:'спутник'/0",
+             "ets:new/2"|_] = lists:reverse(LastCalls),
+            io:format("  last calls ok",[]),
+
+            verify_unicode_stack(Stk),
+            io:format("  unicode stack values ok",[]),
+
 	    {ok,[#ets_table{id="'tab_αβ'",name="'tab_αβ'"}],[]} =
                 crashdump_viewer:ets_tables(Pid),
             io:format("  unicode table name ok",[]),
@@ -653,6 +661,15 @@ special(File,Procs) ->
     end,
     ok.
 
+verify_unicode_stack([{_,{state,Str,Atom,Bin,LongBin}}|_]) ->
+    'unicode_atom_αβ' = Atom,
+    "unicode_string_αβ" = Str,
+    <<"bin αβ"/utf8>> = Bin,
+    <<"long bin αβ - a utf8 binary which can be expanded αβ"/utf8>> = LongBin,
+    ok;
+verify_unicode_stack([_|T]) ->
+    verify_unicode_stack(T).
+
 verify_binaries([H|T1], [H|T2]) ->
     %% Heap binary.
     verify_binaries(T1, T2);
-- 
2.26.2

Places

File 0157-Correct-printing-of-Unicode-atoms-in-crash-dumps.patch of Package erlang (Revision 9e7ef33ae9bf36ae61b08a3391e2483a)

Places