dragonleo/NSIS
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

FileRead in Unicode installers can handle DBCS, conversion output is limited to UCS-2

Browse source 
git-svn-id: https://svn.code.sf.net/p/nsis/code/NSIS/trunk@6399 212acab6-be3b-0410-9dea-997c60f758d6
This commit is contained in:
anders_k 2013-07-27 23:09:54 +00:00
parent 7ce021a376
commit d387a32658
4 changed files with 25 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

2
Docs/src/file.but
View file

@ -26,7 +26,7 @@ Reads a string (ANSI characters) from a file opened with \R{FileOpen}{FileOpen}.
\NsisWarnBlockContainerBegin
\NsisBlockHeaderExeheadU
\#{This is a bug in exehead but it is probably a good idea to document it here...}
The \R{intro-unicode}{Unicode} version can only read text encoded with a single byte character set! The \NsisACPcp is used during the conversion.
DBCS text is supported but conversion output is limited to UCS-2/BMP, surrogate pairs are not supported. The \NsisACPcp is used during the conversion.
\NsisWarnBlockContainerEnd

4
Docs/src/history.but
View file

@ -8,6 +8,10 @@ Released on ?
\S2{} Minor Changes
\b FileRead in Unicode installers can handle DBCS, conversion output is limited to UCS-2.
\b FileRead in Unicode installers now uses the Unicode replacement character (U+FFFD) for invalid characters and not '?'.
\b FileReadByte no longer performs a Unicode conversion on non-ASCII characters
\H{v3.0a1} 3.0 Alpha 1

30
Source/exehead/exec.c
View file

@ -1406,7 +1406,7 @@ static int NSISCALL ExecuteEntry(entry *entry_)
#endif
{
TCHAR *textout=var1;
int rpos=0;
int rpos=0, ungetseek=sizeof(TCHAR);
TCHAR *hptr=var0;
int maxlen=GetIntFromParm(2);
if (maxlen<1) break;
@ -1419,30 +1419,38 @@ static int NSISCALL ExecuteEntry(entry *entry_)
{
TCHAR c;
#ifdef _UNICODE
c=0; // Make sure high byte is 0 for FileReadByte
if (which==EW_FGETS && !parm3)
{
/* BUGBUG:
How is MBTWC supposed to be able to determine the correct WCHAR for a multibyte string when it only has 1 byte to look at?
And what if the multibyte character needs two WCHARs?
*/
char tmpc;
if (!myReadFile(h,&tmpc,1)) break;
if (0==MultiByteToWideChar(CP_ACP, 0, &tmpc, 1, &c, 1)) c = _T('?');
char tmpc[2];
DWORD mbtwcflags=MB_ERR_INVALID_CHARS, cbio;
if (!ReadFile(h,tmpc,2,&cbio,NULL) || !cbio) break;
ungetseek=cbio;
for(;;) // Try to parse as DBCS first, if that fails try again as a single byte
{
// BUGBUG: Limited to UCS-2/BMP, surrogate pairs are not supported.
if (MultiByteToWideChar(CP_ACP,mbtwcflags,tmpc,cbio,&c,1)) break;
c=0xfffd; // Unicode replacement character
// If we read 2 bytes and it was not a DBCS character, we need to seek -1
if (--cbio) SetFilePointer(h,-(--ungetseek),NULL,FILE_CURRENT); else break;
}
}
else
#endif
{
if (!myReadFile(h,&c,1)) break;
// Read 1 TCHAR (FileReadUTF16LE and (Ansi)FileRead) or
// parm3 bytes (FileReadByte and (Unicode)FileReadWord)
if (!myReadFile(h,&c,!parm3 ? sizeof(TCHAR) : sizeof(TCHAR) > 1 ? parm3 : 1)) break;
}
if (parm3)
{
myitoa(textout,(unsigned char)c);
myitoa(textout,(UINT)(_TUCHAR)c);
return 0;
}
if (lc == _T('\r') || lc == _T('\n'))
{
if (lc == c || (c != _T('\r') && c != _T('\n')))
SetFilePointer(h,-((int)(sizeof(c))),NULL,FILE_CURRENT);
SetFilePointer(h,-((int)ungetseek),NULL,FILE_CURRENT);
else
textout[rpos++]=c;
break;

2
Source/script.cpp
View file

@ -5872,7 +5872,7 @@ int CEXEBuild::doCommand(int which_token, LineParser &line)
ent.offsets[0]=GetUserVarIndex(line, 1); // file handle
ent.offsets[1]=GetUserVarIndex(line, 2); // output string
ent.offsets[2]=add_asciistring(_T("1"));
ent.offsets[3]=1;
ent.offsets[3]=2;
if (ent.offsets[0]<0 || ent.offsets[1]<0) PRINTHELP()
SCRIPT_MSG(_T("FileReadWord: %s->%s\n"),line.gettoken_str(1),line.gettoken_str(2));
return add_entry(&ent);