Bug #1011
Mis à jour par Sebastien Jacquot il y a plus de 10 ans
Reported by JCD on Mac OS X 10.9.4: R graphics (WordCloud or CAH) labels are "pixelized".
*Diagnostic 1 - TXM graphics*:
* Question: Can we reproduce the bug?
* Test: produce a CAH from TXM with the following configurations:
** Mac OS X 10.6 + Java 6, 0.65 = no bug
** Mac OS X 10.9 + Java 7, 0.60 + xQuartz 2.7.6 = no bug
** Mac OS X 10.9 + Java 7, 0.67 + xQuartz 2.7.6 = no bug
** Mac OS X 10.9 + Java 7, 0.67 + xQuartz 2.7.7 = no bug
** Remark: Mac OS X 10.9 + Java 7, 0.67 - xQuartz = FAILS, no graphic is displayed. This is expected
** Mac OS X 10.9 - xQuartz = works (verify SJ)
* Conclusion: Impossible to reproduce the bug (MD + SJ) -> JCD must help us to diagnose the bug.
*Diagnostic 2 - JCD R script graphics*:
* Question: Is native R buggy for graphics?
* Test: Run the "bfmprogression.R" script directly from R = no bug
* *Conclusion: ask to set graphic device to 'svg'*
*Diagnostic 3 - JCD TXM SVG graphics*:
* Question: Is the SVG file generated corrupted?
* Test: Get the SVG file generated in the "results" directory then check if it opens well in external editor and in TXM Batik SVG Editor
* Conclusion: nope, the glyphs are just wrong
*Diagnostic 4 - JCD TXM R script graphics*:
* Question: Is TXM R execution for graphics buggy?
* Test: Run "bfmprogression.R" script from TXM
* Conclusion: yes
*Diagnostic 5 - is the bug present in TXM 0.7.5*:
* Question: Has TXM 0.7.6 introduced this bug?
* Test: Reinstall TXM 0.7.5 and compute a CAH graphic
* Conclusion:
*Hypothesis 1*
* see what could go wrong with the 'svg' device of the 'cairo' R package
*WIP*
- define if the bug is only related to the svg() device (SVG device based on cairo graphics) or to all the devices
- to do that, 2 solutions:
1) plot on png, jpg, native window (Quartz) devices in TXM and see if the bug is present
2) plot on svg device in R outside TXM and see if the bug is present
- NOTE: cairo uses X11 on some OS. Therefore on Mac OS X, svg() uses cairo that needs and uses XQuartz
- define if the bug is linked to a font problem
- define if the bug is linked to SVG antialiasing. I got pretty much same results as JCD when setting 'shape-rendering="optimizeSpeed"' attribute of 'svg' tag in an SVG and reopen it in TXM/Batik editor. Need the SVG generated by JCD in 'results' directory to confirm that
*Diagnostic 1 - TXM graphics*:
* Question: Can we reproduce the bug?
* Test: produce a CAH from TXM with the following configurations:
** Mac OS X 10.6 + Java 6, 0.65 = no bug
** Mac OS X 10.9 + Java 7, 0.60 + xQuartz 2.7.6 = no bug
** Mac OS X 10.9 + Java 7, 0.67 + xQuartz 2.7.6 = no bug
** Mac OS X 10.9 + Java 7, 0.67 + xQuartz 2.7.7 = no bug
** Remark: Mac OS X 10.9 + Java 7, 0.67 - xQuartz = FAILS, no graphic is displayed. This is expected
** Mac OS X 10.9 - xQuartz = works (verify SJ)
* Conclusion: Impossible to reproduce the bug (MD + SJ) -> JCD must help us to diagnose the bug.
*Diagnostic 2 - JCD R script graphics*:
* Question: Is native R buggy for graphics?
* Test: Run the "bfmprogression.R" script directly from R = no bug
* *Conclusion: ask to set graphic device to 'svg'*
*Diagnostic 3 - JCD TXM SVG graphics*:
* Question: Is the SVG file generated corrupted?
* Test: Get the SVG file generated in the "results" directory then check if it opens well in external editor and in TXM Batik SVG Editor
* Conclusion: nope, the glyphs are just wrong
*Diagnostic 4 - JCD TXM R script graphics*:
* Question: Is TXM R execution for graphics buggy?
* Test: Run "bfmprogression.R" script from TXM
* Conclusion: yes
*Diagnostic 5 - is the bug present in TXM 0.7.5*:
* Question: Has TXM 0.7.6 introduced this bug?
* Test: Reinstall TXM 0.7.5 and compute a CAH graphic
* Conclusion:
*Hypothesis 1*
* see what could go wrong with the 'svg' device of the 'cairo' R package
*WIP*
- define if the bug is only related to the svg() device (SVG device based on cairo graphics) or to all the devices
- to do that, 2 solutions:
1) plot on png, jpg, native window (Quartz) devices in TXM and see if the bug is present
2) plot on svg device in R outside TXM and see if the bug is present
- NOTE: cairo uses X11 on some OS. Therefore on Mac OS X, svg() uses cairo that needs and uses XQuartz
- define if the bug is linked to a font problem
- define if the bug is linked to SVG antialiasing. I got pretty much same results as JCD when setting 'shape-rendering="optimizeSpeed"' attribute of 'svg' tag in an SVG and reopen it in TXM/Batik editor. Need the SVG generated by JCD in 'results' directory to confirm that