In HTML5 anchors should not be closed when encountering divs, for example, but should be closed when encountering buttons, for example. Many sites that use HTML5-style anchors end up having links displayed with zero-length link texts. The proposed patch correct this behaviour by detecting whether the document is HTML5, then suppressing the close-anchor action in CLOSE_A if it's an HTML5 document. A new macro handles the HTML5-specific cases where anchors are not already always closed.

This also fixes a bug in the tokenizing FSM in etc.c that prevented the !doctype element from being recognized; the fix is necessary because HTML5 detection depends on checking the !doctype element.
This commit is contained in:
Ambrose Li
2020-08-24 23:48:09 -04:00
parent b65f7b243d
commit 48c9ec565d
10 changed files with 72 additions and 6 deletions

3
fm.h
View File

@@ -661,6 +661,7 @@ struct readbuffer {
#endif /* FORMAT_NICE */
#define RB_DEL 0x100000
#define RB_S 0x200000
#define RB_HTML5 0x400000
#define RB_GET_ALIGN(obuf) ((obuf)->flag&RB_ALIGN)
#define RB_SET_ALIGN(obuf,align) {(obuf)->flag &= ~RB_ALIGN; (obuf)->flag |= (align); }
@@ -673,7 +674,7 @@ struct readbuffer {
RB_SET_ALIGN(obuf,(obuf)->flag_stack[--(obuf)->flag_sp]); \
}
/* status flags */
/* state of token scanning finite state machine */
#define R_ST_NORMAL 0 /* normal */
#define R_ST_TAG0 1 /* within tag, just after < */
#define R_ST_TAG 2 /* within tag */