[PATCHv5,03/14] gdb: improve escaping when completing filenames

Message ID 7af24fb43a666452161b1167cd4645a9e9a5fad4.1724173728.git.aburgess@redhat.com
State New
Headers
Series Further filename completion improvements |

Checks

Context Check Description
linaro-tcwg-bot/tcwg_gdb_build--master-aarch64 success Build passed
linaro-tcwg-bot/tcwg_gdb_build--master-arm success Build passed
linaro-tcwg-bot/tcwg_gdb_check--master-aarch64 success Test passed

Commit Message

Andrew Burgess Aug. 20, 2024, 5:10 p.m. UTC
  This improves quoting and escaping when completing filenames for
commands that allow filenames to be quoted and escaped.

I've struggled a bit trying to split this series into chunks.  There's
a lot of dependencies between different parts of the completion
system, and trying to get this working correctly is pretty messy.

This first step is really about implementing 3 readline hooks:

  rl_char_is_quoted_p
    - Is a particular character quoted within readline's input buffer?
  rl_filename_dequoting_function
    - Remove quoting characters from a filename.
  rl_filename_quoting_function
    - Add quoting characters to a filename.

See 'info readline' for full details, but with these hooks connected
up, readline (on behalf of GDB) should do a better job inserting
backslash escapes when completing filenames.

There's still a bunch of stuff that doesn't work after this commit,
mostly around the 'complete' command which of course doesn't go
through readline, so doesn't benefit from all of these new functions
yet, I'll add some of this in a later commit.

Tab completion is now slightly improved though, it is possible to
tab-complete a filename that includes a double or single quote, either
in an unquoted string or within a string surrounded by single or
double quotes, backslash escaping is used when necessary.

There are some additional tests to cover the new functionality.
---
 gdb/completer.c                               | 168 +++++++++++++++++-
 .../gdb.base/filename-completion.exp          |  34 ++++
 2 files changed, 199 insertions(+), 3 deletions(-)
  

Patch

diff --git a/gdb/completer.c b/gdb/completer.c
index 0e49549a643..3ab342dab4f 100644
--- a/gdb/completer.c
+++ b/gdb/completer.c
@@ -214,6 +214,159 @@  noop_completer (struct cmd_list_element *ignore,
 {
 }
 
+/* Return 1 if the character at EINDEX in STRING is quoted (there is an
+   unclosed quoted string), or if the character at EINDEX is quoted by a
+   backslash.  */
+
+static int
+gdb_completer_file_name_char_is_quoted (char *string, int eindex)
+{
+  for (int i = 0; i <= eindex && string[i] != '\0'; )
+    {
+      char c = string[i];
+
+      if (c == '\\')
+	{
+	  /* The backslash itself is not quoted.  */
+	  if (i >= eindex)
+	    return 0;
+	  ++i;
+	  /* But the next character is.  */
+	  if (i >= eindex)
+	    return 1;
+	  if (string[i] == '\0')
+	    return 0;
+	  ++i;
+	  continue;
+	}
+      else if (strchr (rl_completer_quote_characters, c) != nullptr)
+	{
+	  /* This assumes that extract_string_maybe_quoted can handle a
+	     string quoted with character C.  Currently this is true as the
+	     only characters we put in rl_completer_quote_characters are
+	     single and/or double quotes, both of which
+	     extract_string_maybe_quoted can handle.  */
+	  gdb_assert (c == '"' || c == '\'');
+	  const char *tmp = &string[i];
+	  (void) extract_string_maybe_quoted (&tmp);
+	  i = tmp - string;
+
+	  /* Consider any character within the string we just skipped over
+	     as quoted, though this might not be completely correct; the
+	     opening and closing quotes are not themselves quoted.  But so
+	     far this doesn't seem to have caused any issues.  */
+	  if (i >= eindex)
+	    return 1;
+	}
+      else
+	++i;
+    }
+
+  return 0;
+}
+
+/* Removing character escaping from FILENAME.  QUOTE_CHAR is the quote
+   character around FILENAME or the null-character if there is no quoting
+   around FILENAME.  */
+
+static char *
+gdb_completer_file_name_dequote (char *filename, int quote_char)
+{
+  std::string tmp;
+
+  if (quote_char == '\'')
+    {
+      /* There is no backslash escaping within a single quoted string.  In
+	 this case we can just return the input string.  */
+      tmp = filename;
+    }
+  else if (quote_char == '"')
+    {
+      /* Remove escaping from a double quoted string.  */
+      for (const char *input = filename;
+	   *input != '\0';
+	   ++input)
+	{
+	  if (input[0] == '\\'
+	      && input[1] != '\0'
+	      && strchr ("\"\\", input[1]) != nullptr)
+	    ++input;
+	  tmp += *input;
+	}
+    }
+  else
+    {
+      gdb_assert (quote_char == '\0');
+
+      /* Remove escaping from an unquoted string.  */
+      for (const char *input = filename;
+	   *input != '\0';
+	   ++input)
+	{
+	  /* We allow anything to be escaped in an unquoted string.  */
+	  if (*input == '\\')
+	    {
+	      ++input;
+	      if (*input == '\0')
+		break;
+	    }
+
+	  tmp += *input;
+	}
+    }
+
+  return strdup (tmp.c_str ());
+}
+
+/* Apply character escaping to the file name in TEXT.  QUOTE_PTR points to
+   the quote character surrounding TEXT, or points to the null-character if
+   there are no quotes around TEXT.  MATCH_TYPE will be one of the readline
+   constants SINGLE_MATCH or MULTI_MATCH depending on if there is one or
+   many completions.  */
+
+static char *
+gdb_completer_file_name_quote (char *text, int match_type ATTRIBUTE_UNUSED,
+			       char *quote_ptr)
+{
+  std::string str;
+
+  if (*quote_ptr == '\'')
+    {
+      /* There is no backslash escaping permitted within a single quoted
+	 string, so in this case we can just return the input sting.  */
+      str = text;
+    }
+  else if (*quote_ptr == '"')
+    {
+      /* Add escaping for a double quoted filename.  */
+      for (const char *input = text;
+	   *input != '\0';
+	   ++input)
+	{
+	  if (strchr ("\"\\", *input) != nullptr)
+	    str += '\\';
+	  str += *input;
+	}
+    }
+  else
+    {
+      gdb_assert (*quote_ptr == '\0');
+
+      /* Add escaping for an unquoted filename.  */
+      for (const char *input = text;
+	   *input != '\0';
+	   ++input)
+	{
+	  if (strchr (" \t\n\\\"'", *input)
+	      != nullptr)
+	    str += '\\';
+	  str += *input;
+	}
+    }
+
+  return strdup (str.c_str ());
+}
+
 /* Generate filename completions of WORD, storing the completions into
    TRACKER.  This is used for generating completions for commands that
    only accept unquoted filenames as well as for commands that accept
@@ -272,6 +425,7 @@  filename_maybe_quoted_completer_handle_brkchars
     (gdb_completer_file_name_break_characters);
 
   rl_completer_quote_characters = gdb_completer_file_name_quote_characters;
+  rl_char_is_quoted_p = gdb_completer_file_name_char_is_quoted;
 }
 
 /* Complete on filenames.  This is for commands that accepts possibly
@@ -1314,6 +1468,7 @@  complete_line_internal_1 (completion_tracker &tracker,
      completing file names then we can switch to the file name quote
      character set (i.e., both single- and double-quotes).  */
   rl_completer_quote_characters = gdb_completer_expression_quote_characters;
+  rl_char_is_quoted_p = nullptr;
 
   /* Decide whether to complete on a list of gdb commands or on
      symbols.  */
@@ -2209,9 +2364,11 @@  completion_tracker::build_completion_result (const char *text,
   /* Build replacement word, based on the LCD.  */
 
   recompute_lowest_common_denominator ();
-  match_list[0]
-    = expand_preserving_ws (text, end - start,
-			    m_lowest_common_denominator);
+  if (rl_filename_completion_desired)
+    match_list[0] = xstrdup (m_lowest_common_denominator);
+  else
+    match_list[0]
+      = expand_preserving_ws (text, end - start, m_lowest_common_denominator);
 
   if (m_lowest_common_denominator_unique)
     {
@@ -3074,6 +3231,11 @@  _initialize_completer ()
   rl_attempted_completion_function = gdb_rl_attempted_completion_function;
   set_rl_completer_word_break_characters (default_word_break_characters ());
 
+  /* Setup readline globals relating to filename completion.  */
+  rl_filename_quote_characters = " \t\n\\\"'";
+  rl_filename_dequoting_function = gdb_completer_file_name_dequote;
+  rl_filename_quoting_function = gdb_completer_file_name_quote;
+
   add_setshow_zuinteger_unlimited_cmd ("max-completions", no_class,
 				       &max_completions, _("\
 Set maximum number of completion candidates."), _("\
diff --git a/gdb/testsuite/gdb.base/filename-completion.exp b/gdb/testsuite/gdb.base/filename-completion.exp
index 37629bfbf77..c670637ad61 100644
--- a/gdb/testsuite/gdb.base/filename-completion.exp
+++ b/gdb/testsuite/gdb.base/filename-completion.exp
@@ -52,6 +52,9 @@  proc setup_directory_tree {} {
     remote_exec host "touch \"${root}/bb2/dir 2/file 1\""
     remote_exec host "touch \"${root}/bb2/dir 2/file 2\""
 
+    remote_exec host "touch \"${root}/bb1/aa\\\"bb\""
+    remote_exec host "touch \"${root}/bb1/aa'bb\""
+
     return $root
 }
 
@@ -107,6 +110,37 @@  proc run_quoting_and_escaping_tests { root } {
 			"aa cc"
 		    } "" "${qc}" false \
 		    "expand filenames containing spaces"
+
+		test_gdb_complete_multiple "$cmd ${qc}${root}/bb1/" \
+		    "a" "a" {
+			"aa\"bb"
+			"aa'bb"
+		    } "" "${qc}" false \
+		    "expand filenames containing quotes"
+	    } else {
+		set sp "\\ "
+
+		test_gdb_complete_tab_multiple "$cmd ${qc}${root}/aaa/a" \
+		    "a${sp}" {
+			"aa bb"
+			"aa cc"
+		    } false \
+		    "expand filenames containing spaces"
+
+		test_gdb_complete_tab_multiple "$cmd ${qc}${root}/bb1/a" \
+		    "a" {
+			"aa\"bb"
+			"aa'bb"
+		    } false \
+		    "expand filenames containing quotes"
+
+		test_gdb_complete_tab_unique "$cmd ${qc}${root}/bb1/aa\\\"" \
+		    "$cmd ${qc}${root}/bb1/aa\\\\\"bb${qc}" " " \
+		    "expand unique filename containing double quotes"
+
+		test_gdb_complete_tab_unique "$cmd ${qc}${root}/bb1/aa\\'" \
+		    "$cmd ${qc}${root}/bb1/aa\\\\'bb${qc}" " " \
+		    "expand unique filename containing single quote"
 	    }
 	}