This is mostly for Rick, posted here for other's that may be interested.
I worked my way through the new stream library about as far as I can go. I
have a list of the things I found for when you get back to it.
Hopefully, each one will save you a few minutes of investigation. <grin>
Many of the places I wasn't really sure of the proper fix, I just fixed it
enough to continue.
I was using the m_oodcls.rex, trying to get it complete. The list is sort
of in the order I found them.
SysFile::flush() There is the comment about invalidating the buffer, but no
code.
// update the real output position
filePointer += written;
// and invalidate the buffer
I added:
bufferPosition = 0;
bufferedInput = 0;
SysFile::getStreamTypeInfo there is this test:
if ((fileInfo.st_mode & _S_IFREG) != 0)
{
device = true;
transient = true;
}
which marks persistent streams as transient. I think you meant it to be
this test:
if ((fileInfo.st_mode & _S_IFCHR) != 0)
{
device = true;
transient = true;
}
Well, now I forget the order.
StreamInfo::checkEof() There is this code:
// if this is an eof condition, raise that not ready
if (fileInfo.atEof())
{
eof();
}
else
{
// must be an error, so raise that using the file error information
notreadyError();
}
which always generates an error and causes setPosition() to always fail. I
took out the else.
StreamInfo::setPosition() The position value is decremented, than also has
1 subtracted from it in the call to seek().
position--; // convert to system position type
// seek to the target position, if possible. The request position
// is a 1-based character number. We need to convert this into
// a zero-based one before moving.
if (fileInfo.seek(position - 1, SEEK_SET, newPosition))
{
checkEof();
}
I just took out the position--; line
With the rest, I'm even less sure what a good fix is.
bool SysFile::getChar()
In several places in with loops using getChar(), the check is done for a
return of false to fall out of the loop. But getChar() returns true when it
hits end of file.
bool SysFile::getChar(char &ch)
{
size_t len;
return read(&ch, 1, len);
}
read() returns true at eof, and looks like it is intended to return true for
that condition. So, I changed getChar() to only return true if it actually
got a new char:
bool SysFile::getChar(char &ch)
{
size_t len = 0;
return (read(&ch, 1, len) && len == 1);
}
Then there is a basic problem with ~linein() in that the position variables
don't get updated so that linein() in a loop returns the first line over and
over. This one I really kind of hacked. Since the code path I was
producing was going through StreamInfo::readVariableLine(), I did the
position update there.
// If we have a new line character in the last position, we have
// a line. The gets() function has translated crlf sequences into
// single lf characters.
if (buffer[bytesRead - 1] == '\n')
{
charReadPosition += currentLength + bytesRead + (collapsed ? 1 :
0);
lineReadPosition++;
lineReadCharPosition = charReadPosition;
last_op_was_read = true;
return context->NewString(buffer, currentLength + bytesRead -
1);
}
However, since gets() may have collapsed a /r/n into a single byte, I added
in an extra arg that would report if /r/n was collapsed into /n.
fileInfo.gets(readPosition, bufferSize - currentLength, bytesRead,
collapsed)
Not pretty, but it worked. I really wasn't sure what you intended to do
with reads.
The last thing is eof checking when using buffered reads. With a file where
the last bytes only partially fills the buffer, the file is at eof, but
there are still bytes in the buffer that need to be processed.
// hit end of file reading this? This will be the entire line then
if (fileInfo.atEof())
{
...
}
The last read may have read 300 bytes into the buffer. _eof() will return
true, so atEof() returns true. But this line is not really the last line.
I hacked this into:
// hit end of file reading this? This will be the entire line then
if (fileInfo.atEof() && !fileInfo.hasBufferedInput())
{
charReadPosition += currentLength + bytesRead + 1;
lineReadPosition++;
lineReadCharPosition = charReadPosition;
last_op_was_read = true;
return context->NewString(bufferAddress, currentLength +
bytesRead);
}
where I added to SysFile
inline bool hasBufferedInput() { return bufferedInput > bufferPosition;
}
As I said, I didn't mean these to be the correct fixes, just to point out
the areas that need something.
All of the above got m_oodcls.rex working enough to read and write out
several of the input files. But then I get a crash where:
size_t bufferSize;
char *buffer = getDefaultBuffer(bufferSize);
getDefaultBuffer() returns a pointer with the value of: 0x20200a0d
So it looks like the bufferAddress member of StreamInfo has gotten
overwritten. I still get baffled trying to figure those out, so that's
where I quit.
--
Mark Miesfeld
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel
