diff options
| author | Matthias Clasen <mclasen@redhat.com> | 2021-05-09 23:26:32 -0400 |
|---|---|---|
| committer | Matthias Clasen <mclasen@redhat.com> | 2021-05-09 23:29:54 -0400 |
| commit | aa606673d3ae4dae54cef5d7d2de031b8f0b73a1 (patch) | |
| tree | f9e60723a2fded72bb5ef652d4e439cee80c32c7 /pango/break.c | |
| parent | b3ebe097a3042737165ba497dc5b08f235df8167 (diff) | |
| download | pango-aa606673d3ae4dae54cef5d7d2de031b8f0b73a1.tar.gz | |
break: Fix an inconsistency with sentence breaksfix-sentence-breaks
When we retroactively remove a sentence boundary
because of rule SB8, we were not cleaning up the
sencence_start/end markers that have already been
derived from it. This can be seen in urls like
http://www.unicode.org/reports/tr29,
where we don't have any sentence boundaries, but
we leave a stray sentence_start/end at the first
/ after ".org".
Diffstat (limited to 'pango/break.c')
| -rw-r--r-- | pango/break.c | 19 |
1 files changed, 16 insertions, 3 deletions
diff --git a/pango/break.c b/pango/break.c index 6b8e5c01..c34d0512 100644 --- a/pango/break.c +++ b/pango/break.c @@ -306,7 +306,6 @@ pango_default_break (const gchar *text, /* Emoji extended pictographics */ gboolean is_Extended_Pictographic; - wc = next_wc; break_type = next_break_type; @@ -932,7 +931,20 @@ pango_default_break (const gchar *text, prev_prev_SB_type == SB_ATerm_Close_Sp) && IS_OTHER_TERM(prev_SB_type) && SB_type == SB_Lower) - attrs[prev_SB_i].is_sentence_boundary = FALSE; + { + attrs[prev_SB_i].is_sentence_boundary = FALSE; + attrs[prev_SB_i].is_sentence_start = FALSE; + attrs[prev_SB_i].is_sentence_end = FALSE; + last_sentence_start = -1; + for (int j = prev_SB_i - 1; j >= 0; j--) + { + if (attrs[j].is_sentence_boundary) + { + last_sentence_start = j; + break; + } + } + } else if ((prev_SB_type == SB_ATerm || prev_SB_type == SB_ATerm_Close_Sp || prev_SB_type == SB_STerm || @@ -1537,8 +1549,9 @@ pango_default_break (const gchar *text, /* meets space character, move sentence start */ if (last_sentence_start != -1 && last_sentence_start == i - 1 && - attrs[i - 1].is_white) + attrs[i - 1].is_white) { last_sentence_start++; + } } |