Re: Re: About GNU Ghostscript 707

Yannis,

Could please provide a small PostScript file with DSC comments 
that triggers this problem?  I don't have AI10.

Adding nulls to the end of the buffer should not be needed.

Russell

On 11 Sep 2003 at 0:53, Yannis Calotychos wrote:

> Ghostscript : all versions
> file        : dscparse.c
> function    : dsc_scan_data()
> system      : Apple PowerBook G4 800/Mac OS X 1.2.6
> remarks     :
> the buffer kept by ghostscript is getting refilled when reading the ps 
> file.
> occasionally the new data combined with the data already in the buffer 
> causes the parser to enter an infinite loop.
> for example the buffer may look like:
> ABCDEFGH
> 
> after token H is parsed, the buffer may be refilled to
> IJKLEFGH
> 
> the new buffer can cause the parser to enter an infinite loop whereby 
> all tokens in "IJKLEFGH" are parsed.
> a simple fix that zeros trailing parts of the buffer (to IJKL0000) 
> fixes this issue.
> this issue crashes GS with most sample files provided by the latest 
> Adobe Illustrator installation (AI10 for Mac OS X).
> 
> I hope this makes sense,
> Please review, and possibly incorporate the code shown below (change is 
> bracketed in comments /* Y.Calotychos - iView Multimedia Ltd. */)
> 
> We have a few more changes regarding functions mac_find_font_family() 
> and mac_get_font_encoding() that coould be used to load the correct 
> fonts. however we are still trying to understand how gs calculates the 
> metrics of strings as, altgough our corrections work in therms of the 
> font that is used, the character spacing/kerning is not correct.
> I would be glad to send sample code.
> 
> 
> ------------------------------------
> 
> 
> int
> dsc_scan_data(CDSC *dsc, const char *data, int length)
> {
>      int bytes_read;
>      int code = 0;
> 
>      if (dsc == NULL)
> 	return CDSC_ERROR;
> 
>      if (dsc->id == CDSC_NOTDSC)
> 	return CDSC_NOTDSC;
>      dsc->id = CDSC_OK;
>      if (dsc->eof)
> 	return CDSC_OK;	/* ignore */
> 
>      if (length == 0) {
> 	/* EOF, so process what remains */
> 	dsc->eof = TRUE;
>      }
> 
>      do {
> 	if (dsc->id == CDSC_NOTDSC)
> 	    break;
> 
> 	if (length != 0) {
> 		/* Y.Calotychos - iView Multimedia Ltd. */
> 		
> 		/* OLD CODE --------------
> 	    if (dsc->data_length > CDSC_DATA_LENGTH/2) {
> 		memmove(dsc->data, dsc->data + dsc->data_index,
> 		    dsc->data_length - dsc->data_index);
> 		dsc->data_offset += dsc->data_index;
> 		dsc->data_length -= dsc->data_index;
> 		dsc->data_index = 0;
> 	    } END OF OLD CODE -------- */
> 	
> 	
> 	    if (dsc->data_length > CDSC_DATA_LENGTH/2)
> 	    {
> 	    	// copy the remaining data to the top of the buffer.
> 	    	// we really nead to clear this array as parsers can get confused 
> as to where
> 	    	// this line breaks
> 	    	long l = dsc->data_length - dsc->data_index;
> 			memmove(dsc->data, dsc->data + dsc->data_index, l);
> 			memset(dsc->data+l, 0, CDSC_DATA_LENGTH-l);
> 
> 
> 			dsc->data_offset += dsc->data_index;
> 			dsc->data_length -= dsc->data_index;
> 			dsc->data_index = 0;
> 	    }
> 		/* Y.Calotychos - iView Multimedia Ltd. */
> 
> 
> 	    /* append to buffer */
> 	    bytes_read = min(length, (int)(CDSC_DATA_LENGTH - 
> dsc->data_length));
> 	    memcpy(dsc->data + dsc->data_length, data, bytes_read);
> 	    dsc->data_length += bytes_read;
> 	    data += bytes_read;
> 	    length -= bytes_read;
> 	}
> 	if (dsc->scan_section == scan_none) {
> 	    code = dsc_scan_type(dsc);
> 	    if (code == CDSC_NEEDMORE) {
> 		/* need more characters before we can identify type */
> 		code = CDSC_OK;
> 		break;
> 	    }
> 	    dsc->id = code;
> 	}
> 
>          if (code == CDSC_NOTDSC) {
> 	    dsc->id = CDSC_NOTDSC;
> 	    break;
> 	}
> 
> 	while ((code = dsc_read_line(dsc)) > 0) {
> 	    if (dsc->id == CDSC_NOTDSC)
> 		break;
> 	    if (dsc->file_length &&
> 		(dsc->data_offset + dsc->data_index > dsc->file_length)) {
> 		/* have read past end of where we need to parse. */
> 		return CDSC_OK;	/* ignore */
> 	    }
> 	    if (dsc->doseps_end &&
> 		(dsc->data_offset + dsc->data_index > dsc->doseps_end)) {
> 		/* have read past end of DOS EPS PostScript section */
> 		return CDSC_OK;	/* ignore */
> 	    }
> 	    if (dsc->eof)
> 		return CDSC_OK;
> 	    if (dsc->skip_document)
> 		continue;	/* embedded document */
> 	    if (dsc->skip_lines)
> 		continue;	/* embedded lines */
> 	    if (IS_DSC(dsc->line, "%%BeginData:"))
> 		continue;
> 	    if (IS_DSC(dsc->line, "%%BeginBinary:"))
> 		continue;
> 	    if (IS_DSC(dsc->line, "%%EndDocument"))
> 		continue;
> 	    if (IS_DSC(dsc->line, "%%EndData"))
> 		continue;
> 	    if (IS_DSC(dsc->line, "%%EndBinary"))
> 		continue;
> 
> 	    do {
> 		switch (dsc->scan_section) {
> 		    case scan_comments:
> 			code = dsc_scan_comments(dsc);
> 			break;
> 		    case scan_pre_preview:
> 		    case scan_preview:
> 			code = dsc_scan_preview(dsc);
> 			break;
> 		    case scan_pre_defaults:
> 		    case scan_defaults:
> 			code = dsc_scan_defaults(dsc);
> 			break;
> 		    case scan_pre_prolog:
> 		    case scan_prolog:
> 			code = dsc_scan_prolog(dsc);
> 			break;
> 		    case scan_pre_setup:
> 		    case scan_setup:
> 			code = dsc_scan_setup(dsc);
> 			break;
> 		    case scan_pre_pages:
> 		    case scan_pages:
> 			code = dsc_scan_page(dsc);
> 			break;
> 		    case scan_pre_trailer:
> 		    case scan_trailer:
> 			code = dsc_scan_trailer(dsc);
> 			break;
> 		    case scan_eof:
> 			code = CDSC_OK;
> 			break;
> 		    default:
> 			/* invalid state */
> 			code = CDSC_ERROR;
> 		}
> 		/* repeat if line is start of next section */
> 	    } while (code == CDSC_PROPAGATE);
> 
> 	    /* if DOS EPS header not complete, ask for more */
> 	    if (code == CDSC_NEEDMORE) {
> 		code = CDSC_OK;
> 		break;
> 	    }
> 	    if (code == CDSC_NOTDSC) {
> 		dsc->id = CDSC_NOTDSC;
> 		break;
> 	    }
> 	}
>      } while (length != 0);
> 
>      return (code < 0) ? code : dsc->id;
> }
> 
> 
> 
> -------------------------------------
> yannis calotychos ph.d.
> www.iview-multimedia.com
> 
> _______________________________________________
> gs-code-review mailing list
> [email protected]
> http://www.ghostscript.com/mailman/listinfo/gs-code-review



Russell Lang                   [email protected]
Ghostgum Software Pty Ltd      http://www.ghostgum.com.au/