On 2009-01-05 10:33+0100 Arjen Markus wrote: > [Andrew wrote] >> Just for the record, your changes make no difference to me. This is on >> Ubuntu Hardy x86 32-bit with tcl 8.4. I still have rounding differences >> for examples 3, 11 and 20 as before. I can't test example 21 as it is >> not enable with tcl 8.4. Will try to upgrade to tcl 8.5 and test again. >> > > I should have mentioned I have been doing these tests with Tcl 8.5. > The difference between Tcl 8.4 and Tcl 8.5 that is of importance here > is that in 8.5 the parameter that determines the precision for > converting numbers to strings (tcl_precision) is set to 17 by default, > making an exact reconversion possible. > > (The Tcl interface still uses the old string-based mechanism, hence > tcl_precision is important. I intend to change that to the Tcl_Obj > mechanism.) > > Example 11 shows very small differences: several numbers like 2384 > instead of 2383 in the PostScript output. Perhaps I can get rid of > these too, but I gave priority to the others (examples 15 and 16 are > fine then?). > > I will have a look at example 3. It is a new one to me. > > Example 20 is causing me a slight headache :(. I have checked just > about everything I could think of: > - All expressions are braced > - The file is read correctly - all values are identical to the > values I see in the C example > - The PostScript file differs in the way the gray colour is > determined, but I have not been able to determine why that > would be the case. plscmap1n and plscmap1l are called with > identical arguments. > > I will have to descent further into the library functions to > determine the cause.
Hi Arjen: Here is what I now (revision 9253) get with Tcl-8.5 on Debian testing. tcl Missing examples : 19 Differing postscript output : 11 20 Missing stdout : Differing stdout : That's a substantial improvement from what I got before for Tcl-8.5. In particular some large differences that could not be analyzed by ndiff for two examples (15 and 16?) have been eliminated. Good work! The remaining differences for Tcl-8.5 are just for examples 11 and 20. The ndiff result confirms what you said about the small differences for example 11. softw...@raven> ndiff -relerr 0.9 x11[ct].psc |head 6c6 < %%CreationDate: Mon Jan 5 10:32:05 2009 --- field 5 > %%CreationDate: Mon Jan 5 10:33:42 2009 ### Maximum relative error in matching lines = 7.79e-04 at line 33792 field 16 Here is the ndiff summary for example 20. softw...@raven> ~/ndiff/install/bin/ndiff -relerr 1 x20[ct].psc 6c6 < %%CreationDate: Mon Jan 5 10:32:06 2009 --- field 5 > %%CreationDate: Mon Jan 5 10:33:44 2009 ### Maximum relative error in matching lines = 1.0e+00 at line 255608 field 1 To see the first two examples of that type of large relative error ndiff should be run as follows: softw...@raven> ~/ndiff/install/bin/ndiff -relerr 0.9 x20[ct].psc |head -12 6c6 < %%CreationDate: Mon Jan 5 10:32:06 2009 --- field 5 > %%CreationDate: Mon Jan 5 10:33:44 2009 255608c255608 < 0.0039 0.0039 0.0039 C Z --- field 1 relative error 1.0e+00 > 0.0000 0.0000 0.0000 C Z 307524c307524 < 0.0039 0.0039 0.0039 C Z --- field 1 relative error 1.0e+00 > 0.0000 0.0000 0.0000 C Z These are RGB PostScript colour settings so this confirms your result that there is an important Tcl-8.5 colour issue with example 20. In sum, the results I get on my Linux Tcl-8.5 (64-bit) platform seem to confirm everything you have found on your Windows Tcl-8.5 platform including the good results you have now achieved for all examples other than 11 and 20. That sort of cross-platform confirmation is reassuring, and I hope it encourages you to go on to solve the issues for example 11 and 20. Alan __________________________ Alan W. Irwin Astronomical research affiliation with Department of Physics and Astronomy, University of Victoria (astrowww.phys.uvic.ca). Programming affiliations with the FreeEOS equation-of-state implementation for stellar interiors (freeeos.sf.net); PLplot scientific plotting software package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of Linux Links project (loll.sf.net); and the Linux Brochure Project (lbproject.sf.net). __________________________ Linux-powered Science __________________________ ------------------------------------------------------------------------------ _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
