Open tklib tickets: Not reproduced

Triage of all 71 open and Pending tickets of tklib (bugs, patches and feature requests) on 2026-11-15: whether a bug still exists in tklib trunk with Tk 9.1 and 8.6, whether a patch is still needed, and whether a request has been implemented, superseded or become outdated. 25 prototype patches with regression tests; rows marked patch have one. None is committed.

Not reproduced 5

⬜214b46fbad2025widget::dateentry does not work on WindowsNOT-REPRODUCED (Windows 11, Tk 9.1.1 and 8.6.18, tklib trunk, widget::dateentry 0.99)

Open ticket 214b46fbad · widget :: dateentry · Bug · filed 2025-02-02

Full verdict

NOT-REPRODUCED (Windows 11, Tk 9.1.1 and 8.6.18, tklib trunk, widget::dateentry 0.99)

  • Module widget (widget::dateentry) · Bug
  • Tested Tk 9.1.1, 8.6.18 on X11: both a simulated click and post map the dropdown on top. Repro script: win/214b46fbad.tcl (prints mapped state, stacking order and the window under the pointer, then a verdict).
  • Cause unknown. Likely candidates: the popdown is an overrideredirect toplevel made transient at creation time, with no -topmost. ttk::combobox sets wm transient just before posting and sets -topmost 1 on win32 (see the NOTE-WM-TRANSIENT comment in Tk's combobox.tcl). It may also be ttk::globalGrab on <Map> combined with tkwait visibility.
  • Prototype none (needs Windows to verify)
  • Suggested ticket action Not reproducible on Linux. Run win/214b46fbad.tcl on Windows. If it reproduces, change post to set wm transient and wm attributes -topmost 1 just before deiconify, as ttk::combobox does.
Windows VM, run 1
  • Verdict NOT-REPRODUCED (Windows 11, Tk 9.1.1 and 8.6.18, tklib trunk, widget::dateentry 0.99)
  • Tested win/214b46fbad.tcl: a simulated click on the entry and post both map the dropdown, it is on top of the stacking order (. .d.__drop) and the window under the pointer is the calendar. The display is scaled (350%), so the dropdown is 738x436.
  • Note the events were synthesized with event generate; a real mouse click was not tried.
  • Suggested ticket action Could not reproduce with tklib trunk and Tk 8.6.18 or 9.1.1 on Windows 11. Please try a current tklib, and say which Windows and Tk version you use if it still happens.
⬜009b09029f2022Axis labels are not always "nice"patchNOT-REPRODUCED for the reported case: the XY plot, strip chart, histogram, isometric plot, bar charts, log plots and TX plot with {121.45 121.51 0.…

Open ticket 009b09029f · plotchart · Bug · filed 2022-01-26

Full verdict

NOT-REPRODUCED for the reported case: the XY plot, strip chart, histogram, isometric plot, bar charts, log plots and TX plot with {121.45 121.51 0.02} / {0.0 2.0 0.2} all give clean labels, because these axes already format with %.12g (since 2010/2011). The 2022 legend change a8df88b405 is unrelated. However, I found two related cases that do reproduce on Tk 9.1 and 8.6:

  • Module plotchart · Bug
    • a reversed axis {2.0 0.0 -0.2} labels zero as 2.77555756156e-16;
    • the polar plot radial labels show 0.6000000000000001, 1.7999999999999998, and so on.
  • Tested Tk 9.1, 8.6: canvas text items for 14 plot types.
  • Cause plotaxis.tcl, axis loops in DrawYaxis, DrawRightaxis, DrawXaxis and DrawXtext: the snap-to-zero check tests the tick position $y/$x instead of the label value $yt/$xt. In DrawPolarAxes, $rad is used as the label text without formatting.
  • Prototype 009b09029f.patch (5 code lines plus tests Plotchart-6.1 and Plotchart-6.2 in plotchart.test). The plotchart suite (46 tests) passes on 9.1 and 8.6.
  • Suggested ticket action The normal axes already use %.12g. Fixed the remaining cases: the zero label on reversed axes and the polar radial labels. Ask the reporter which plot type showed 121.49999999 if it persists; otherwise close as Fixed.
⬜f4b93930512020Error on ctext when C comments enabledNOT-REPRODUCED (3000 random fuzz inputs to ctext::comments on Tk 9.1 and 8.6: quotes, escapes, comment markers, newlines, non-BMP chars, elided r…

Open ticket f4b9393051 · ctext · Patch · filed 2020-04-22

Full verdict

NOT-REPRODUCED (3000 random fuzz inputs to ctext::comments on Tk 9.1 and 8.6: quotes, escapes, comment markers, newlines, non-BMP chars, elided ranges; no error). The patch still applies; the code is unchanged since 2013.

  • Module ctext · Patch
  • Tested Tk 9.1, 8.6 fuzz of ctext::comments with comments enabled.
  • Cause in stock ctext, $win search -count length is passed through uplevel 1 to the inner text widget, so length is always set when an index is returned. The reported error most likely comes from a wrapper or renamed widget command that does not forward the -count variable to the caller's frame.
  • Prototype none (not reproducible, so no regression test). The 3-line defensive check from the ticket is harmless.
  • Suggested ticket action Could not reproduce with current ctext on Tk 8.6/9.1. Ask for a script that reproduces it, e.g. whether the ctext command is wrapped. Apply the defensive info exists length check if desired, otherwise close as Works For Me.
⬜a697f07a382015Infinite map/unmap of autoscroll widgetsNOT-REPRODUCED. The attachment TestAutoscroll.zip (blob 20950e04…) is missing from the repository, so the original script cannot be run. I tried …

Open ticket a697f07a38 · autoscroll · Bug · filed 2015-04-01

Full verdict

NOT-REPRODUCED. The attachment TestAutoscroll.zip (blob 20950e04…) is missing from the repository, so the original script cannot be run. I tried text widgets with vertical and horizontal autoscroll scrollbars, -wrap none/char/word, contents sized right at the fit/no-fit boundary, fixed and propagating toplevels, and resizing the toplevel in 3 px steps over 200–520 px × 5 heights. This was done with no WM and under openbox. No oscillation: at most 2–4 maps per change, never continuous.

  • Module autoscroll · Bug
  • Tested Tk 9.1, 8.6 (Xvfb, no WM and openbox).
  • Cause unknown. autoscroll::map freezes the toplevel geometry to prevent exactly this kind of loop. A known theoretical source is the text widget's asynchronous line-metric updates while the scrollbar toggles.
  • Prototype none (no reproducer)
  • Suggested ticket action The attached example is no longer available in the repository and I could not reproduce a map/unmap loop with Tk 8.6/9.1. Please attach the script again (inline), otherwise close as Works For Me.
⬜2499978fff2009Infinite bug on widget::scrolledwindowNOT-REPRODUCED (X11: ttk::treeview with items in an auto-scrolled scrolledwindow, 35 window sizes; and a synthetic widget whose content reflows whe…

Open ticket 2499978fff · tklib · Bug · filed 2009-01-11

Full verdict

NOT-REPRODUCED (X11: ttk::treeview with items in an auto-scrolled scrolledwindow, 35 window sizes; and a synthetic widget whose content reflows when a scrollbar appears)

  • Module widget (widget::scrolledwindow) · Bug
  • Tested Tk 9.1.1, 8.6.18. No recursion error. The lock in _set_scroll lets scrollbars be added but not removed while locked, so the layout settles.
  • Cause the reporter said it happens only on Windows XP. The traceback is from BWidget Tree::_update_scrollregion, not ttk::treeview. scrollw.tcl _set_scroll has not changed since 2007 (19442b0fab). A small weakness remains: a nested call clears sb(lock) while the outer call is still inside update idletasks. The proposed one-line patch (if {$sb(auto) && !$sb(lock)}) avoids that, but it can also leave a needed scrollbar hidden.
  • Prototype none (cannot reproduce; the platform is gone)
  • Suggested ticket action Cannot reproduce with Tk 8.6/9.x on Linux, and the report is specific to Windows XP with BWidget Tree. Close as "Out of Date" unless someone can reproduce it on a current Windows.