Qt5 5.6 regression: conversion from QVariant to QString is broken

Bug #1620173 reported by Mathieu Pellerin
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
qtbase-opensource-src (Ubuntu)
Confirmed
Undecided
Unassigned

Bug Description

A regression was introduced in Qt5 5.5 - fixed in 5.7 - whereas conversion from QVariant to QString is broken for double and float values (you can see more details on this here: https://bugreports.qt.io/browse/QTBUG-47192).

The patch fixing this issue is the following: http://code.qt.io/cgit/qt/qtbase.git/patch/src/corelib/tools/qlocale.h?id=726fed0d67013cbfac7921d3d4613ca83406fb0f

Would it be possible to apply this patch against the package (based on Qt5 5.6.1) Ubuntu 16.10 will ship in October? It's a pretty ugly bug, which needs addressing. Hopefully the Qt devs can eventually merge this patch onto their 5.6 LTS branch.

description: updated
Revision history for this message
Mathieu Pellerin (nirvn-asia) wrote :

Ok, while trying to create a debdiff (with the help of Simon Q on IRC, thanks :) ), I realized the above is effectively part 2, and part 1 is this commit:

Part 1: https://codereview.qt-project.org/#/c/127924/12//ALL,unified
Part 2: https://codereview.qt-project.org/#/c/127924/12//ALL,unified (same as the .patch link above)

This fixes QTBUG-44039, QTBUG-43885, and QTBUG-47192.

Revision history for this message
Timo Jyrinki (timo-jyrinki) wrote :

Thanks for the report and finding both of the commits!

Changed in qtbase-opensource-src (Ubuntu):
assignee: nobody → Timo Jyrinki (timo-jyrinki)
status: New → Confirmed
Revision history for this message
Timo Jyrinki (timo-jyrinki) wrote :

I think you meant https://codereview.qt-project.org/#/c/104737/32/ for patch 1 :D

Revision history for this message
Timo Jyrinki (timo-jyrinki) wrote :

The patches do not apply to 5.6 as is, although it doesn't look too complicated failing.

Revision history for this message
Mathieu Pellerin (nirvn-asia) wrote :

Timo, yes sorry, you've got the proper link :)

Revision history for this message
Timo Jyrinki (timo-jyrinki) wrote :

Also https://codereview.qt-project.org/#/c/139057/ is needed. Next time, full debdiff with working and rebased patches please :)

Revision history for this message
Timo Jyrinki (timo-jyrinki) wrote :

Unfortunately fixing this would be causing true regressions in at least ktexteditor and pyqt5:

https://requests.ci-train.ubuntu.com/static/britney/landing-1920/yakkety/excuses.html

https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac/autopkgtest-yakkety-ci-train-ppa-service-landing-1920/yakkety/amd64/k/ktexteditor/20160914_111856@/log.gz
https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac/autopkgtest-yakkety-ci-train-ppa-service-landing-1920/yakkety/amd64/p/pyqt5/20160914_140056@/log.gz

I'm not sure if rebuilding qtxmlpatterns would be enough or what other kind of problems there could be lurking. Since 16.10 release is coming closer, I think the set of these three patches (that remove a symbol among else) is too risky. The fix would eventually be in Qt 5.7 in Ubuntu 17.04, unless of course a less invasive fix could be had for 16.10.

Changed in qtbase-opensource-src (Ubuntu):
assignee: Timo Jyrinki (timo-jyrinki) → nobody
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.