// ---------------------------------------------------------------------
namespace APT {
namespace String {
-std::string Strip(const std::string &s)
+std::string Strip(const std::string &str)
{
- size_t start = s.find_first_not_of(" \t\n");
- // only whitespace
- if (start == string::npos)
+ // ensure we have at least one character
+ if (str.empty() == true)
+ return str;
+
+ char const * const s = str.c_str();
+ size_t start = 0;
+ for (; isspace(s[start]) != 0; ++start)
+ ; // find the first not-space
+
+ // string contains only whitespaces
+ if (s[start] == '\0')
return "";
- size_t end = s.find_last_not_of(" \t\n");
- return s.substr(start, end-start+1);
+
+ size_t end = str.length() - 1;
+ for (; isspace(s[end]) != 0; --end)
+ ; // find the last not-space
+
+ return str.substr(start, end - start + 1);
}
bool Endswith(const std::string &s, const std::string &end)
return (s.substr(s.size() - end.size(), s.size()) == end);
}
+bool Startswith(const std::string &s, const std::string &start)
+{
+ if (start.size() > s.size())
+ return false;
+ return (s.substr(0, start.size()) == start);
+}
+
}
}
/*}}}*/
then returns the result. Several varients on true/false are checked. */
int StringToBool(const string &Text,int Default)
{
- char *End;
- int Res = strtol(Text.c_str(),&End,0);
- if (End != Text.c_str() && Res >= 0 && Res <= 1)
+ char *ParseEnd;
+ int Res = strtol(Text.c_str(),&ParseEnd,0);
+ // ensure that the entire string was converted by strtol to avoid
+ // failures on "apt-cache show -a 0ad" where the "0" is converted
+ const char *TextEnd = Text.c_str()+Text.size();
+ if (ParseEnd == TextEnd && Res >= 0 && Res <= 1)
return Res;
// Check for positives
In particular: this reads blocks from the input until it believes
that it's run out of input text. Each block is terminated by a
- double newline ('\n' followed by '\n'). As noted below, there is a
- bug in this code: it assumes that all the blocks have been read if
- it doesn't see additional text in the buffer after the last one is
- parsed, which will cause it to lose blocks if the last block
- coincides with the end of the buffer.
+ double newline ('\n' followed by '\n').
*/
bool ReadMessages(int Fd, vector<string> &List)
{
char Buffer[64000];
- char *End = Buffer;
// Represents any left-over from the previous iteration of the
// parse loop. (i.e., if a message is split across the end
// of the buffer, it goes here)
string PartialMessage;
-
- while (1)
- {
- int Res = read(Fd,End,sizeof(Buffer) - (End-Buffer));
+
+ do {
+ int const Res = read(Fd, Buffer, sizeof(Buffer));
if (Res < 0 && errno == EINTR)
continue;
-
- // Process is dead, this is kind of bad..
+
+ // process we read from has died
if (Res == 0)
return false;
-
+
// No data
- if (Res < 0 && errno == EAGAIN)
+ if (Res < 0 && (errno == EAGAIN || errno == EWOULDBLOCK))
return true;
if (Res < 0)
return false;
-
- End += Res;
-
- // Look for the end of the message
- for (char *I = Buffer; I + 1 < End; I++)
+
+ // extract the message(s) from the buffer
+ char const *Start = Buffer;
+ char const * const End = Buffer + Res;
+
+ char const * NL = (char const *) memchr(Start, '\n', End - Start);
+ if (NL == NULL)
{
- if (I[1] != '\n' ||
- (I[0] != '\n' && strncmp(I, "\r\n\r\n", 4) != 0))
- continue;
-
- // Pull the message out
- string Message(Buffer,I-Buffer);
- PartialMessage += Message;
-
- // Fix up the buffer
- for (; I < End && (*I == '\n' || *I == '\r'); ++I);
- End -= I-Buffer;
- memmove(Buffer,I,End-Buffer);
- I = Buffer;
-
- List.push_back(PartialMessage);
- PartialMessage.clear();
+ // end of buffer: store what we have so far and read new data in
+ PartialMessage.append(Start, End - Start);
+ Start = End;
}
- if (End != Buffer)
- {
- // If there's text left in the buffer, store it
- // in PartialMessage and throw the rest of the buffer
- // away. This allows us to handle messages that
- // are longer than the static buffer size.
- PartialMessage += string(Buffer, End);
- End = Buffer;
- }
else
- {
- // BUG ALERT: if a message block happens to end at a
- // multiple of 64000 characters, this will cause it to
- // terminate early, leading to a badly formed block and
- // probably crashing the method. However, this is the only
- // way we have to find the end of the message block. I have
- // an idea of how to fix this, but it will require changes
- // to the protocol (essentially to mark the beginning and
- // end of the block).
- //
- // -- dburrows 2008-04-02
- return true;
- }
+ ++NL;
+
+ if (PartialMessage.empty() == false && Start < End)
+ {
+ // if we start with a new line, see if the partial message we have ended with one
+ // so that we properly detect records ending between two read() runs
+ // cases are: \n|\n , \r\n|\r\n and \r\n\r|\n
+ // the case \r|\n\r\n is handled by the usual double-newline handling
+ if ((NL - Start) == 1 || ((NL - Start) == 2 && *Start == '\r'))
+ {
+ if (APT::String::Endswith(PartialMessage, "\n") || APT::String::Endswith(PartialMessage, "\r\n\r"))
+ {
+ PartialMessage.erase(PartialMessage.find_last_not_of("\r\n") + 1);
+ List.push_back(PartialMessage);
+ PartialMessage.clear();
+ while (NL < End && (*NL == '\n' || *NL == '\r')) ++NL;
+ Start = NL;
+ }
+ }
+ }
+
+ while (Start < End) {
+ char const * NL2 = (char const *) memchr(NL, '\n', End - NL);
+ if (NL2 == NULL)
+ {
+ // end of buffer: store what we have so far and read new data in
+ PartialMessage.append(Start, End - Start);
+ break;
+ }
+ ++NL2;
+
+ // did we find a double newline?
+ if ((NL2 - NL) == 1 || ((NL2 - NL) == 2 && *NL == '\r'))
+ {
+ PartialMessage.append(Start, NL2 - Start);
+ PartialMessage.erase(PartialMessage.find_last_not_of("\r\n") + 1);
+ List.push_back(PartialMessage);
+ PartialMessage.clear();
+ while (NL2 < End && (*NL2 == '\n' || *NL2 == '\r')) ++NL2;
+ Start = NL2;
+ }
+ NL = NL2;
+ }
+
+ // we have read at least one complete message and nothing left
+ if (PartialMessage.empty() == true)
+ return true;
if (WaitFd(Fd) == false)
return false;
- }
+ } while (true);
}
/*}}}*/
// MonthConv - Converts a month string into a number /*{{{*/
// ---------------------------------------------------------------------
/* This is used in decoding the 256bit encoded fixed length fields in
tar files */
-bool Base256ToNum(const char *Str,unsigned long &Res,unsigned int Len)
+bool Base256ToNum(const char *Str,unsigned long long &Res,unsigned int Len)
{
if ((Str[0] & 0x80) == 0)
return false;
}
}
/*}}}*/
+// Base256ToNum - Convert a fixed length binary to a number /*{{{*/
+// ---------------------------------------------------------------------
+/* This is used in decoding the 256bit encoded fixed length fields in
+ tar files */
+bool Base256ToNum(const char *Str,unsigned long &Res,unsigned int Len)
+{
+ unsigned long long Num;
+ bool rc;
+
+ rc = Base256ToNum(Str, Num, Len);
+ Res = Num;
+ if (Res != Num)
+ return false;
+
+ return rc;
+}
+ /*}}}*/
// HexDigit - Convert a hex character into an integer /*{{{*/
// ---------------------------------------------------------------------
/* Helper for Hex2Num */
projects / apt.git / blobdiff