Known Problems in PGPLOT

Version 5.2.0

General Problems

Text. There is no way to display a subscript of a superscript or a superscript of a subscript. This is due to a design flaw in PGPLOT. I don't think there is any way that I can fix this problem with the current escape codes \u and \d: there is no way to distinguish a \d meaning end-of-superscript from a \d meaning start-of-subscript-of-superscript. I will try to address this in a future version of PGPLOT, if I can think of a way of doing so while retaining backwards compatibility. I think that I will have to introduce new bracketting delimiters, like HTML <sup>...</sup> and </sub>...</sub> or TeX's ^{...} and _{...}.
There is no printed manual yet!
CPGPLOT (C binding):
- Routines that require the name of a subroutine as an argument (PGCONX, PGFUNX, PGFUNY, PGFUNT, etc.) are not handled by the C binding.
- CPGPLOT requires an ANSI C compiler.

Problems with Specific Subroutines

PGBIN: the first and last bins (in CENTER mode) or the last bin (in non-center mode) are plotted incorrectly; a redesign is needed, because the arguments to PGBIN don't provide enough information to do the job.
PGBOX (and PGENV): if the axis range is small compared with the absolute values of the endpoints, an integer overflow can occur (VMS) or the axis can be labeled incorrectly (UNIX). Example: range 2200.0 to 2200.01.
PGCONF: has some algorithmic problems noted in the subroutine synopsis.
PGCONL: with arrays larger than 100×100 pixels, the routine does not place labels precisely as advertised. This is a design flaw. It arises because the contouring routines do a segment of the array at a time (not more than 100×100). This breaks continuous contours and interferes with the labeling sequence. I tried to mitigate this in 5.0.3 by increasing the segment from 50x50 to 100×100, but this only helps a bit. I'll try to do something about this in a later version. Ideally I would like to allocate a temporary work array equal in size to the array to be contoured, but that is not very portable. I also need to make labeling and blanking (PGCONL and PGCONT) work together; at present you can have either one but not both. PGCONL implicitly requires that PGCINT be greater than PGCMIN; this is an unnecessary restriction.
PGNUMB: this routine does not always honor a request for decimal format (1) but will use an exponential format instead.
PGQDT: note that some PGPLOT installations include ``stub'' drivers that should not be called. PGQDT returns a blank TYPE string and TLEN=0 for such drivers. If you use PGQDT in a program, you should test for this condition before using the information returned by PGQDT.

System Problems

OSF1 (sys_osf1):
- All the Fortran drivers that use the %VAL() mechanism for passing an address fail under DEC UNIX (OSF/1); these drivers will work if you change the declarations of certain pointer variables from INTEGER to INTEGER*8. In several of the drivers, I have included comments to indicate what should be changed.
- Cpgdemo does not link on OSF/1: error message cpgdemo.o: main: multiply defined; Unresolved: MAIN__ The best solution to this is to use the flag -nofor_main on the f77 command used to link the program (thanks to David Terrett).
Silicon Graphics (sys_irix):
- In IRIX 6.2, the library names have changed. If the installation fails to find the X11 library, Edit the .conf file in the pgplot/sys_irix directory and change
```
   LIBS="-lX11_s"
```
  to
```
   LIBS="-lX11"
```
  and then rerun makemake.
- A user has reported that PGNUMB does not generate correct code for numbers in exponential format. This affects axes drawn with PGBOX etc. when labels require a multiplying power of 10. This appears to be a due to a bug in the SGI Fortran compiler. It can apparently be fixed by specifying -backslash when compiling the pgplot library. Edit the .conf file and change
```
FFLAGC="-u -O2"
```
  to
```
FFLAGC="-u -O2 -backslash"
```
- The installation procedure for cpgdemo fails. ld reports unresolved __main. The command is
```
f77 -o cpgdemo cpgdemo.o -L`pwd` -lcpgplot -lpgplot -lX11_s
```
  A work-around is to link using cc instead:
```
 
cc -o cpgdemo cpgdemo.o -L`pwd` -lcpgplot -lpgplot -lX11_s -lftn -lm
```
VMS (sys_vms):
- Installation procedure should compile only required drivers.
- Installation procedure assumes that a C compiler (CC) is available. It should be modified to skip the C compilations when this is not true.

Device drivers

GIdriv (GIF driver):
- GIF uses the LZW compression system which is patented by Unisys; a license may be required to use this driver.
- Uses space-inefficient algorithm.
- It might be nicer to insert page-number before the .gif extension instead of after, but what if there is no extension?
HPGL drivers:
- In some porgrams, the HPGL plot will be filled with many horizontal lines. I think this is caused by using PGERAS, which is trying to fill the page with black, when it should be filling in the background color. As the HPGL has no erase capability, the PGERAS call should really be ignored. Unfortunately I have no easy way to test the HPGL driver; if someone would like to work with me on this problem, or can point me to a viewer for HPGL files for UNIX machines, please contact me.
HJdriv, LJdriv (Hewlett-Packard PCL)
- I have no way to test these drivers, and I am not at all sure that they work on any systems except VMS. I don't know whether to recommend one or the other. I would appreciate feedback.
GOdriv, VBdriv, ZEdriv
- These drivers require a routine grge00 that is not available on UNIX systems. They should be rewritten to improve portability. Do not select these drivers on UNIX systems.
PSdriv (PostScript handler):
- The round terminals of thick lines are not clipped at the edge of the window (this bug affects all handlers that do hardware thick lines).
TTdriv (/TK4100):
- set-color-representation doesn't work correctly; colors are not reset on exit (should they be?).
VTdriv (VT125 handler):
- Cursor input does not work correctly. Unfortunately I do not have any good way to test this.
X2DRIV, PGDISP: Note that this driver is no longer supported and the following problems will not be fixed; use the X-Window driver (/XSERV) instead.
- "pgdisp -lineColor 2" crashes immediately.
- figdisp_comm.c fails to compile in systems that lack "values.h" (e.g., ConvexOS 10.1).
- The driver tries to start up "figdisp" instead of "pgdisp"; you can change this in the code (figdisp_comm.c) if you wish.
- PGDISP cannot be compiled with DECC on VAX/VMS (not AXP).
- Compilation of pgdisp on AXP makes warning messages:
```
   %LINK-W-MULDEF, symbol DECC$NTOHS multiply defined
        in module DECC$SHR file SYS$COMMON:[SYSLIB]DECC$SHR.EXE;1
   %LINK-W-MULDEF, symbol DECC$HTONS multiply defined
        in module DECC$SHR file SYS$COMMON:[SYSLIB]DECC$SHR.EXE;1
```
XWdriv (X Window driver):
- Does not compile with some non-ANSI compilers (e.g., Convex?) that do not have "string.h".
- Does not support the old XEdriv methods of specifying a label for the PGPLOT window title bar.
VTdriv (Dec REGIS driver):
This driver has problems on VMS. There is a special VMS variant of this driver in the [PGPLOT.DRIVERS] directory. Rename file VTDRIV-VMS.F to VTDRIV.F (replacing the file of that name, which is designed for UNIX), and recompile the library. This driver will accept a device specification of the form "/VT" to display on the terminal, or "file/VT" to create a disk file that you can display with TYPE. The driver still has some problems, however: (1) it is limited to 4 colors, even if your REGIS device can display more; (2) I think the cursor doesn't work on most REGIS devices, although I believe it works on a real VT125 terminal (a rare beast indeed); (3) if you direct output to a file, the driver still thinks the output is an interacrtive device and issues a prompt for each new page. I do not plan to fix any of these problems, as REGIS is now an obsolete graphics language.
Tektronix drivers (TTdriv, TFdriv) and disk files:
If you want your graphics to display in an XTERM window, you should use a device specification "/XTERM" (or possibly "terminal/XTERM" if you have acces to a 'terminal' other than the one you are running the program in. Device type "/TEK", designed for a real Tektronix terminal, does not generate the code required to switch an XTERM into Tektronix mode. You can use /TEK with an XTERM if you first switch the terminal into Tektronix mode by hand, using the control-middle-mouse-button menu.
Device spec "file/TFILE" produces a disk file in Tektronix format. There are two points to note: (1) it does not include the code to switch an XTERM, so you can only display it in an XTERM if the XTERM is already in Tektronix mode. (2) It uses the Tektronix extended (high-resolution) addressing mode, and apparently XTERM doesn't recognize this, so the plot is slightly garbled.
There are many variants of the Tektronix format, and this variety is exacerbated on VMS where you cannot use the same code to write to a terminal as you would use to write a disk file.
I will try to address this problem in a future version of PGPLOT. Basically, I need to modify the I/O routines used by the Tektronix drivers so that they can be used to write a disk file as well as send output to a terminal. This would give you the option of a disk file with any of the Tektronix variants (/TEK, /XTERM, etc.).