diff options
author | toma <toma@283d02a7-25f6-0310-bc7c-ecb5cbfe19da> | 2009-11-25 17:56:58 +0000 |
---|---|---|
committer | toma <toma@283d02a7-25f6-0310-bc7c-ecb5cbfe19da> | 2009-11-25 17:56:58 +0000 |
commit | 460c52653ab0dcca6f19a4f492ed2c5e4e963ab0 (patch) | |
tree | 67208f7c145782a7e90b123b982ca78d88cc2c87 /libkmime/kmime_codec_uuencode.cpp | |
download | tdepim-460c52653ab0dcca6f19a4f492ed2c5e4e963ab0.tar.gz tdepim-460c52653ab0dcca6f19a4f492ed2c5e4e963ab0.zip |
Copy the KDE 3.5 branch to branches/trinity for new KDE 3.5 features.
BUG:215923
git-svn-id: svn://anonsvn.kde.org/home/kde/branches/trinity/kdepim@1054174 283d02a7-25f6-0310-bc7c-ecb5cbfe19da
Diffstat (limited to 'libkmime/kmime_codec_uuencode.cpp')
-rw-r--r-- | libkmime/kmime_codec_uuencode.cpp | 240 |
1 files changed, 240 insertions, 0 deletions
diff --git a/libkmime/kmime_codec_uuencode.cpp b/libkmime/kmime_codec_uuencode.cpp new file mode 100644 index 000000000..2c6152a95 --- /dev/null +++ b/libkmime/kmime_codec_uuencode.cpp @@ -0,0 +1,240 @@ +/* -*- c++ -*- + kmime_codec_uuencode.cpp + + This file is part of KMime, the KDE internet mail/usenet news message library. + Copyright (c) 2002 Marc Mutz <mutz@kde.org> + + KMime is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License, version 2, as + published by the Free Software Foundation. + + KMime is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this library; if not, write to the Free Software + Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + + In addition, as a special exception, the copyright holders give + permission to link the code of this library with any edition of + the Qt library by Trolltech AS, Norway (or with modified versions + of Qt that use the same license as Qt), and distribute linked + combinations including the two. You must obey the GNU General + Public License in all respects for all of the code used other than + Qt. If you modify this file, you may extend this exception to + your version of the file, but you are not obligated to do so. If + you do not wish to do so, delete this exception statement from + your version. +*/ + +#include "kmime_codec_uuencode.h" + +#include <kdebug.h> + +#include <cassert> + +using namespace KMime; + +namespace KMime { + + +class UUDecoder : public Decoder { + uint mStepNo; + uchar mAnnouncedOctetCount; // (on current line) + uchar mCurrentOctetCount; // (on current line) + uchar mOutbits; + bool mLastWasCRLF : 1; + bool mSawBegin : 1; // whether we already saw ^begin... + uint mIntoBeginLine : 3; // count #chars we compared against "begin" 0..5 + bool mSawEnd : 1; // whether we already saw ^end... + uint mIntoEndLine : 2; // count #chars we compared against "end" 0..3 + + void searchForBegin( const char* & scursor, const char * const send ); + +protected: + friend class UUCodec; + UUDecoder( bool withCRLF=false ) + : Decoder( withCRLF ), mStepNo(0), + mAnnouncedOctetCount(0), mCurrentOctetCount(0), + mOutbits(0), mLastWasCRLF(true), + mSawBegin(false), mIntoBeginLine(0), + mSawEnd(false), mIntoEndLine(0) {} + +public: + virtual ~UUDecoder() {} + + bool decode( const char* & scursor, const char * const send, + char* & dcursor, const char * const dend ); + // ### really needs no finishing??? + bool finish( char* & /*dcursor*/, const char * const /*dend*/ ) { return true; } +}; + + + +Encoder * UUCodec::makeEncoder( bool ) const { + return 0; // encoding not supported +} + +Decoder * UUCodec::makeDecoder( bool withCRLF ) const { + return new UUDecoder( withCRLF ); +} + + + /********************************************************/ + /********************************************************/ + /********************************************************/ + + + +void UUDecoder::searchForBegin( const char* & scursor, const char * const send ) { + static const char begin[] = "begin\n"; + static const uint beginLength = 5; // sic! + + assert( !mSawBegin || mIntoBeginLine > 0 ); + + while ( scursor != send ) { + uchar ch = *scursor++; + if ( ch == begin[mIntoBeginLine] ) { + if ( mIntoBeginLine < beginLength ) { + // found another char + ++mIntoBeginLine; + if ( mIntoBeginLine == beginLength ) + mSawBegin = true; // "begin" complete, now search the next \n... + } else /* mIntoBeginLine == beginLength */ { + // found '\n': begin line complete + mLastWasCRLF = true; + mIntoBeginLine = 0; + return; + } + } else if ( mSawBegin ) { + // OK, skip stuff until the next \n + } else { + kdWarning() << "UUDecoder: garbage before \"begin\", resetting parser" + << endl; + mIntoBeginLine = 0; + } + } + +} + + +// uuencoding just shifts all 6-bit octets by 32 (SP/' '), except NUL, +// which gets mapped to 0x60 +static inline uchar uuDecode( uchar c ) { + return ( c - ' ' ) // undo shift and + & 0x3F; // map 0x40 (0x60-' ') to 0... +} + + +bool UUDecoder::decode( const char* & scursor, const char * const send, + char* & dcursor, const char * const dend ) +{ + // First, check whether we still need to find the "begin" line: + if ( !mSawBegin || mIntoBeginLine != 0 ) + searchForBegin( scursor, send ); + // or if we are past the end line: + else if ( mSawEnd ) { + scursor = send; // do nothing anymore... + return true; + } + + while ( dcursor != dend && scursor != send ) { + uchar ch = *scursor++; + uchar value; + + // Check whether we need to look for the "end" line: + if ( mIntoEndLine > 0 ) { + static const char end[] = "end"; + static const uint endLength = 3; + + if ( ch == end[mIntoEndLine] ) { + ++mIntoEndLine; + if ( mIntoEndLine == endLength ) { + mSawEnd = true; + scursor = send; // shortcut to the end + return true; + } + continue; + } else { + kdWarning() << "UUDecoder: invalid line octet count looks like \"end\" (mIntoEndLine = " << mIntoEndLine << " )!" << endl; + mIntoEndLine = 0; + // fall through... + } + } + + // Normal parsing: + + // The first char of a line is an encoding of the length of the + // current line. We simply ignore it: + if ( mLastWasCRLF ) { + // reset char-per-line counter: + mLastWasCRLF = false; + mCurrentOctetCount = 0; + + // try to decode the chars-on-this-line announcement: + if ( ch == 'e' ) // maybe the beginning of the "end"? ;-) + mIntoEndLine = 1; + else if ( ch > 0x60 ) + {} // ### invalid line length char: what shall we do?? + else if ( ch > ' ' ) + mAnnouncedOctetCount = uuDecode( ch ); + else if ( ch == '\n' ) + mLastWasCRLF = true; // oops, empty line + + continue; + } + + // try converting ch to a 6-bit value: + if ( ch > 0x60 ) + continue; // invalid char + else if ( ch > ' ' ) + value = uuDecode( ch ); + else if ( ch == '\n' ) { // line end + mLastWasCRLF = true; + continue; + } else + continue; + + // add the new bits to the output stream and flush full octets: + switch ( mStepNo ) { + case 0: + mOutbits = value << 2; + break; + case 1: + if ( mCurrentOctetCount < mAnnouncedOctetCount ) + *dcursor++ = (char)(mOutbits | value >> 4); + ++mCurrentOctetCount; + mOutbits = value << 4; + break; + case 2: + if ( mCurrentOctetCount < mAnnouncedOctetCount ) + *dcursor++ = (char)(mOutbits | value >> 2); + ++mCurrentOctetCount; + mOutbits = value << 6; + break; + case 3: + if ( mCurrentOctetCount < mAnnouncedOctetCount ) + *dcursor++ = (char)(mOutbits | value); + ++mCurrentOctetCount; + mOutbits = 0; + break; + default: + assert( 0 ); + } + mStepNo = (mStepNo + 1) % 4; + + // check whether we ran over the announced octet count for this line: + kdWarning( mCurrentOctetCount == mAnnouncedOctetCount + 1 ) + << "UUDecoder: mismatch between announced (" + << mAnnouncedOctetCount << ") and actual line octet count!" << endl; + + } + + // return false when caller should call us again: + return (scursor == send); +} // UUDecoder::decode() + + +} // namespace KMime |