Changeset 178099 in webkit

Timestamp:

Jan 7, 2015 9:18:20 PM (9 years ago)

Author:

commit-queue@webkit.org

Message:

The ASCII decoding for non ASCII character is incorrect if this character comes after going through the fast decoding code path and before the end of the text by less than a machine word size of characters.
​https://bugs.webkit.org/show_bug.cgi?id=140173.

Patch by Said Abou-Hallawa <​sabouhallawa@apple.com> on 2015-01-07
Reviewed by Darin Adler.

Source/WebCore:

Tests: fast/encoding/char-after-fast-path-ascii-decoding.html.

• platform/text/TextCodecLatin1.cpp:

(WebCore::TextCodecLatin1::decode):
This function has a bug when it goes through the fast decoding code path. After copying
one or more all ASCII MachineWords from source to the destination, the following byte
is copied as is from the source to the destination even if it is non ASCII byte. This
causes the decoded bytes to be incorrect. The fix is to ensure that the current byte
is still ASCII after exiting the fast decoding code path.

LayoutTests:

• fast/encoding/char-after-fast-path-ascii-decoding-expected.txt: Added.
• fast/encoding/char-after-fast-path-ascii-decoding.html: Added.

Ensures when an non ASCII character comes after a machine word, whose bytes are all
ASCII characters, is decoded correctly.

Location:

trunk

Files:

• LayoutTests/ChangeLog (modified) (1 diff)
• LayoutTests/fast/encoding/char-after-fast-path-ascii-decoding-expected.txt (added)
• LayoutTests/fast/encoding/char-after-fast-path-ascii-decoding.html (added)
• Source/WebCore/ChangeLog (modified) (1 diff)
• Source/WebCore/platform/text/TextCodecLatin1.cpp (modified) (2 diffs)

trunk/LayoutTests/ChangeLog

@@ +1 @@
+2015-01-07  Said Abou-Hallawa  <sabouhallawa@apple.com>
+
+        The ASCII decoding for non ASCII character is incorrect if this character comes after going through the fast decoding code path and before the end of the text by less than a machine word size of characters.
+        https://bugs.webkit.org/show_bug.cgi?id=140173.
+
+        Reviewed by Darin Adler.
+
+        * fast/encoding/char-after-fast-path-ascii-decoding-expected.txt: Added.
+        * fast/encoding/char-after-fast-path-ascii-decoding.html: Added.
+        Ensures when an non ASCII character comes after a machine word, whose bytes are all
+        ASCII characters, is decoded correctly.
+
 2015-01-07  Shivakumar JM  <shiva.jm@samsung.com>
 

trunk/Source/WebCore/ChangeLog

@@ +1 @@
+2015-01-07  Said Abou-Hallawa  <sabouhallawa@apple.com>
+
+        The ASCII decoding for non ASCII character is incorrect if this character comes after going through the fast decoding code path and before the end of the text by less than a machine word size of characters.
+        https://bugs.webkit.org/show_bug.cgi?id=140173.
+
+        Reviewed by Darin Adler.
+
+        Tests: fast/encoding/char-after-fast-path-ascii-decoding.html.
+
+        * platform/text/TextCodecLatin1.cpp:
+        (WebCore::TextCodecLatin1::decode):
+        This function has a bug when it goes through the fast decoding code path. After copying
+        one or more all ASCII MachineWords from source to the destination, the following byte
+        is copied as is from the source to the destination even if it is non ASCII byte. This
+        causes the decoded bytes to be incorrect. The fix is to ensure that the current byte
+        is still ASCII after exiting the fast decoding code path.
+ 
 2015-01-07  Shivakumar JM  <shiva.jm@samsung.com>
 

trunk/Source/WebCore/platform/text/TextCodecLatin1.cpp

@@ -147 +147 @@
                 if (source == end)
                     break;
+
+                // *source may not be ASCII anymore if source moves inside the loop of the fast code path
+                if (!isASCII(*source))
+                    goto useLookupTable;
             }
             *destination = *source;
@@ -198 +202 @@
                 if (source == end)
                     break;
+
+                // *source may not be ASCII anymore if source moves inside the loop of the fast code path
+                if (!isASCII(*source))
+                    goto useLookupTable16;
             }
             *destination16 = *source;