From 94f9cf81ed3e156bd372a3cac249974d1acb4e1d Mon Sep 17 00:00:00 2001 From: "jan.nijtmans" Date: Sun, 24 Jun 2018 20:26:27 +0000 Subject: Fix "string tolower" and friends for handling unpaired surrogates correctly. Also add test-cases for those situations. Various typo's in comments. --- doc/Utf.3 | 5 ++++- generic/tclCmdMZ.c | 3 +++ generic/tclDisassemble.c | 2 +- generic/tclExecute.c | 14 ++++++++++---- generic/tclParse.c | 1 + generic/tclUtf.c | 21 +++++++++++++++------ tests/utf.test | 15 +++++++++++++++ 7 files changed, 49 insertions(+), 12 deletions(-) diff --git a/doc/Utf.3 b/doc/Utf.3 index 160575b..922fd81 100644 --- a/doc/Utf.3 +++ b/doc/Utf.3 @@ -132,7 +132,10 @@ represent one Unicode character in the UTF-8 representation. .PP \fBTcl_UniCharToUtf\fR stores the character \fIch\fR as a UTF-8 string in starting at \fIbuf\fR. The return value is the number of bytes stored -in \fIbuf\fR. +in \fIbuf\fR. If ch is an upper surrogate (range U+D800 - U+DBFF), then +the return value will be 0 and nothing will be stored. If you still +want to produce UTF-8 output for it (even though knowing it's an illegal +code-point on its own), just call \fBTcl_UniCharToUtf\fR again using ch = -1. .PP \fBTcl_UtfToUniChar\fR reads one UTF-8 character starting at \fIsrc\fR and stores it as a Tcl_UniChar in \fI*chPtr\fR. The return value is the diff --git a/generic/tclCmdMZ.c b/generic/tclCmdMZ.c index d64299e..0bd6cb4 100644 --- a/generic/tclCmdMZ.c +++ b/generic/tclCmdMZ.c @@ -1447,6 +1447,9 @@ StringIndexCmd( char buf[4]; length = Tcl_UniCharToUtf(ch, buf); + if (!length) { + length = Tcl_UniCharToUtf(-1, buf); + } Tcl_SetObjResult(interp, Tcl_NewStringObj(buf, length)); } } diff --git a/generic/tclDisassemble.c b/generic/tclDisassemble.c index e9aaec4..a0d1258 100644 --- a/generic/tclDisassemble.c +++ b/generic/tclDisassemble.c @@ -894,7 +894,7 @@ PrintSourceToObj( Tcl_AppendPrintfToObj(appendObj, "\\U%08x", ch); i += 10; } else -#elif TCL_UTF_MAX > 3 +#else /* If len == 0, this means we have a char > 0xffff, resulting in * TclUtfToUniChar producing a surrogate pair. We want to output * this pair as a single Unicode character. diff --git a/generic/tclExecute.c b/generic/tclExecute.c index fda50b2..82de752 100644 --- a/generic/tclExecute.c +++ b/generic/tclExecute.c @@ -4964,7 +4964,7 @@ TEBCresume( /* Decode index value operands. */ - /* + /* assert ( toIdx != TCL_INDEX_AFTER); * * Extra safety for legacy bytecodes: @@ -5223,9 +5223,15 @@ TEBCresume( * but creating the object as a string seems to be faster in * practical use. */ - - length = (ch != -1) ? Tcl_UniCharToUtf(ch, buf) : 0; - objResultPtr = Tcl_NewStringObj(buf, length); + if (ch == -1) { + objResultPtr = Tcl_NewObj(); + } else { + length = Tcl_UniCharToUtf(ch, buf); + if (!length) { + length = Tcl_UniCharToUtf(-1, buf); + } + objResultPtr = Tcl_NewStringObj(buf, length); + } } TRACE_APPEND(("\"%s\"\n", O2S(objResultPtr))); diff --git a/generic/tclParse.c b/generic/tclParse.c index 581270c..00b83a1 100644 --- a/generic/tclParse.c +++ b/generic/tclParse.c @@ -981,6 +981,7 @@ TclParseBackslash( } count = Tcl_UniCharToUtf(result, dst); if (!count) { + /* Special case for handling upper surrogates. */ count = Tcl_UniCharToUtf(-1, dst); } return count; diff --git a/generic/tclUtf.c b/generic/tclUtf.c index ed10ab2..c8292a2 100644 --- a/generic/tclUtf.c +++ b/generic/tclUtf.c @@ -218,7 +218,7 @@ Tcl_UniCharToUtfDString( { const Tcl_UniChar *w, *wEnd; char *p, *string; - int oldLength; + int oldLength, len = 1; /* * UTF-8 string length in bytes will be <= Unicode string length * 4. @@ -231,9 +231,18 @@ Tcl_UniCharToUtfDString( p = string; wEnd = uniStr + uniLength; for (w = uniStr; w < wEnd; ) { - p += Tcl_UniCharToUtf(*w, p); + if (!len && ((*w & 0xFC00) != 0xDC00)) { + /* Special case for handling upper surrogates. */ + p += Tcl_UniCharToUtf(-1, p); + } + len = Tcl_UniCharToUtf(*w, p); + p += len; w++; } + if (!len) { + /* Special case for handling upper surrogates. */ + p += Tcl_UniCharToUtf(-1, p); + } Tcl_DStringSetLength(dsPtr, oldLength + (p - string)); return string; @@ -899,7 +908,7 @@ Tcl_UtfToUpper( * char to dst if its size is <= the original char. */ - if (bytes < TclUtfCount(upChar)) { + if ((bytes < TclUtfCount(upChar)) || ((upChar & 0xF800) == 0xD800)) { memcpy(dst, src, (size_t) bytes); dst += bytes; } else { @@ -962,7 +971,7 @@ Tcl_UtfToLower( * char to dst if its size is <= the original char. */ - if (bytes < TclUtfCount(lowChar)) { + if ((bytes < TclUtfCount(lowChar)) || ((lowChar & 0xF800) == 0xD800)) { memcpy(dst, src, (size_t) bytes); dst += bytes; } else { @@ -1022,7 +1031,7 @@ Tcl_UtfToTitle( #endif titleChar = Tcl_UniCharToTitle(titleChar); - if (bytes < TclUtfCount(titleChar)) { + if ((bytes < TclUtfCount(titleChar)) || ((titleChar & 0xF800) == 0xD800)) { memcpy(dst, src, (size_t) bytes); dst += bytes; } else { @@ -1046,7 +1055,7 @@ Tcl_UtfToTitle( lowChar = Tcl_UniCharToLower(lowChar); } - if (bytes < TclUtfCount(lowChar)) { + if ((bytes < TclUtfCount(lowChar)) || ((lowChar & 0xF800) == 0xD800)) { memcpy(dst, src, (size_t) bytes); dst += bytes; } else { diff --git a/tests/utf.test b/tests/utf.test index 67a6778..e820359 100644 --- a/tests/utf.test +++ b/tests/utf.test @@ -158,6 +158,12 @@ test utf-8.3 {Tcl_UniCharAtIndex: index > 0} { test utf-8.4 {Tcl_UniCharAtIndex: index > 0} { string index \u4e4e\u25a\xff\u543 2 } "\uff" +test utf-8.5 {Tcl_UniCharAtIndex: upper surrogate} { + string index \ud842 0 +} "\ud842" +test utf-8.5 {Tcl_UniCharAtIndex: lower surrogate} { + string index \udc42 0 +} "\udc42" test utf-9.1 {Tcl_UtfAtIndex: index = 0} { string range abcd 0 2 @@ -263,6 +269,9 @@ test utf-11.4 {Tcl_UtfToUpper} { test utf-11.5 {Tcl_UtfToUpper Georgian (new in Unicode 11)} { string toupper \u10d0\u1c90 } \u1c90\u1c90 +test utf-11.6 {Tcl_UtfToUpper low/high surrogate)} { + string toupper \udc24\ud824 +} \udc24\ud824 test utf-12.1 {Tcl_UtfToLower} { string tolower {} @@ -279,6 +288,9 @@ test utf-12.4 {Tcl_UtfToLower} { test utf-12.5 {Tcl_UtfToLower Georgian (new in Unicode 11)} { string tolower \u10d0\u1c90 } \u10d0\u10d0 +test utf-12.6 {Tcl_UtfToUpper low/high surrogate)} { + string tolower \udc24\ud824 +} \udc24\ud824 test utf-13.1 {Tcl_UtfToTitle} { string totitle {} @@ -298,6 +310,9 @@ test utf-13.5 {Tcl_UtfToTitle Georgian (new in Unicode 11)} { test utf-13.6 {Tcl_UtfToTitle Georgian (new in Unicode 11)} { string totitle \u1c90\u10d0 } \u1c90\u10d0 +test utf-13.7 {Tcl_UtfToTitle low/high surrogate)} { + string totitle \udc24\ud824 +} \udc24\ud824 test utf-14.1 {Tcl_UtfNcasecmp} { string compare -nocase a b -- cgit v0.12