Skip to contentSkip to conversation

Fixed an issue which contained extra unnecessary code

PATCH In review sushmaunnibhavi 3 participants 3 messages Mar 10, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fixed an issue which contained extra unnecessary code

From 5a6c233c6bf0a35aca000b32b9e936a34532900a Mon Sep 17 00:00:00 2001
From: sushmaunnibhavi <[email protected]>
Date: Sun, 10 Mar 2019 19:37:33 +0530
Subject: [GSOC][PATCH] Fixed an issue which contained extra unnecessary code

---
Since '\n' and '\0' are char_len==1,it is not necessary to check if the char_len<=1.

 compat/regex/regexec.c | 5 -----
 1 file changed, 5 deletions(-)
compat/regex/regexec.c +0 -5
index 1b5d89fd5e..df62597531 100644
@@ -3799,11 +3799,6 @@ check_node_accept_bytes (const re_dfa_t *dfa, int node_idx,
   char_len = re_string_char_size_at (input, str_idx);
   if (node->type == OP_PERIOD)
     {
-      if (char_len <= 1)
-	return 0;
-      /* FIXME: I don't think this if is needed, as both '\n'
-	 and '\0' are char_len == 1.  */
-      /* '.' accepts any one character except the following two cases.  */
       if ((!(dfa->syntax & RE_DOT_NEWLINE) &&
 	   re_string_byte_at (input, str_idx) == '\n') ||
 	  ((dfa->syntax & RE_DOT_NOT_NULL) &&
Signature · 1 line
2.17.1
#1 lore raw
Quoted 10 lines · sushmaunnibhavi

From 5a6c233c6bf0a35aca000b32b9e936a34532900a Mon Sep 17 00:00:00 2001
From: sushmaunnibhavi <[email protected]>
Date: Sun, 10 Mar 2019 19:37:33 +0530
Subject: [GSOC][PATCH] Fixed an issue which contained extra unnecessary code

---
Since '\n' and '\0' are char_len==1,it is not necessary to check if the char_len<=1.

 compat/regex/regexec.c | 5 -----
 1 file changed, 5 deletions(-)

It doesn't look like the patch is formatted correctly. I think that the explanation line ("Since '\n' and '\0' are...") should be above the line that contains your "Signed-off-by: ..." and there should be a blank line between those two lines.

Also we ask for an author name in the "From: ..." header that looks like "Firstname Lastname". A simple way to do that would be to make it match the name in your "Signed-off-by: ...".

Quoted 6 lines
compat/regex/regexec.c +0 -0
index 1b5d89fd5e..df62597531 100644
@@ -3799,11 +3799,6 @@ check_node_accept_bytes (const re_dfa_t *dfa, int node_idx,
   char_len = re_string_char_size_at (input, str_idx);

compat/regex/regexec.c check_node_accept_bytes (const re_dfa_t *dfa, int node_idx, · line 3804

Quoted 10 lines
compat/regex/regexec.c +0 -5
   if (node->type == OP_PERIOD)
     {
-      if (char_len <= 1)
-       return 0;
-      /* FIXME: I don't think this if is needed, as both '\n'
-        and '\0' are char_len == 1.  */
-      /* '.' accepts any one character except the following two cases.  */
       if ((!(dfa->syntax & RE_DOT_NEWLINE) &&
           re_string_byte_at (input, str_idx) == '\n') ||
          ((dfa->syntax & RE_DOT_NOT_NULL) &&

The code looks like:

char_len = re_string_char_size_at (input, str_idx); if (node->type == OP_PERIOD)

    {
      if (char_len <= 1)
    return 0;
      /* FIXME: I don't think this if is needed, as both '\n'
     and '\0' are char_len == 1.  */
      /* '.' accepts any one character except the following two cases.  */
      if ((!(dfa->syntax & RE_DOT_NEWLINE) &&
       re_string_byte_at (input, str_idx) == '\n') ||
      ((dfa->syntax & RE_DOT_NOT_NULL) &&
       re_string_byte_at (input, str_idx) == '\0'))
    return 0;
      return char_len;
    }

If the byte at re_string_byte_at (input, str_idx) is indeed '\n' or '\0', then yeah probably char_len == 1 so the current code should return 0 just before the code below the FIXME comment is reached. So in this case the 2 checks below the FIXME comment are useless because they check re_string_byte_at (input, str_idx) == '\n') and re_string_byte_at (input, str_idx) == '\0' which cannot happen.

So I would say that the right fix would be to remove those 2 checks, not to remove the if (char_len <= 1) check above the FIXME comment.

Also note that I used "probably" when I wrote "then yeah probably char_len == 1", because I think it is worth checking what re_string_char_size_at() actually does before being too much confident...

#2 lore raw
In reply to sushmaunnibhavi
Quoted 1 line

Subject: [GSOC][PATCH] Fixed an issue which contained extra unnecessary code

Commit messages (and titles) should always be in the imperative mood. The title in particular should be a short description of what the patch is doing, and should give meaningful information to people reading the output of 'git log --oneline'. Also we usually prefix the title with the area of the code we're touching.

The title above is very generic, and thus not very useful to future readers. Something like

compat/regex: remove unnecessary if

would give future readers a bit more context.

Quoted 4 lines · sushmaunnibhavi

From 5a6c233c6bf0a35aca000b32b9e936a34532900a Mon Sep 17 00:00:00 2001
From: sushmaunnibhavi <[email protected]>
Date: Sun, 10 Mar 2019 19:37:33 +0530
Subject: [GSOC][PATCH] Fixed an issue which contained extra unnecessary code

The four lines above don't need to be included here, as they are already defined by the email headers.

The one line that may be useful here is "From: sushmaunnibhavi <[email protected]>". That should be your full name, and match the name in the Signed-off-by line. Usually people just set that up in the "From" header in the email, but if you are unable to configure your mailer like that, including a "From: Sushma Unnibhavi <[email protected]>" before the rest of the commit message also works.

Quoted 3 lines

---
Since '\n' and '\0' are char_len==1,it is not necessary to check if the char_len<=1.

This explanation of why this is a good patch should be included in the commit message before your Signed-off-by. If we want to apply this change (which I don't think we want for the reasons stated below), I think this needs a bit of a deeper analysis in the commit message, to convince readers this is the correct thing to do.

Also please wrap commit messages at 72 characters.

compat/regex/regexec.c

Quoted 4 lines
 compat/regex/regexec.c | 5 -----
 1 file changed, 5 deletions(-)
compat/regex/regexec.c +0 -0

This file in compat/ was directly imported from upstream gawk. We generally don't patch this type of imported file, to make updates from upstream easier, unless there is an actual fix from upstream that needs to be fixed that's not going to be fixed upstream.

As this change at best removes a redundant 'if' (I can't comment on the correctness, as I'm not familar with this code), so it is not worth changing this file in our codebase.

Quoted 18 lines

index 1b5d89fd5e..df62597531 100644

compat/regex/regexec.c +0 -5
@@ -3799,11 +3799,6 @@ check_node_accept_bytes (const re_dfa_t *dfa, int node_idx,
   char_len = re_string_char_size_at (input, str_idx);
   if (node->type == OP_PERIOD)
     {
-      if (char_len <= 1)
-	return 0;
-      /* FIXME: I don't think this if is needed, as both '\n'
-	 and '\0' are char_len == 1.  */
-      /* '.' accepts any one character except the following two cases.  */
       if ((!(dfa->syntax & RE_DOT_NEWLINE) &&
 	   re_string_byte_at (input, str_idx) == '\n') ||
 	  ((dfa->syntax & RE_DOT_NOT_NULL) &&
Signature · 1 line
2.17.1