PyQt5 status

[Steve Borho]

Hello,

I’ve just built binaries for Hg 4.2.3 and Hg+Thg 4.3.1 and am now looking into what it will take to migrate the Windows Thg packages to Qt5.

Other than the dearth of PyQt5 binary packages, which I discussed earlier, the main blocker is now Kdiff3. There hasn’t been a Kdiff3 release in three years, and while they claim Qt5 support there have been no releases since they added the feature, and no-one else is making Windows binaries that I’ve found.

Options:

1 - we un-bundle kdiff3 and tell people to download and install it separately, or some other merge tool
2 - someone builds and tests it against Qt-5.6 and sends me new binaries
3 - we package Kdiff3 statically linked with Qt4


Another problem I’ve encountered on Ubuntu Linux is that with PyQt5 the QSci widgets all have incorrect cursor positioning, it appears the cursor is positioned based on an integer font width but the font itself has a fractional (floating point) width. IE: when the cursor is at the end of 14 monospace characters, it is rendered between the 13th and 14th. It’s pretty disorienting. Can anyone else reproduce this? I’ve tried a few workarounds to resolve the problem, but was not successful.

—
Steve

[Michał Sznajder]

Hi

On Fri, Aug 11, 2017 at 5:13 PM, Steve Borho <st...@borho.org> wrote:
>
> Hello,
>
> I’ve just built binaries for Hg 4.2.3 and Hg+Thg 4.3.1 and am now looking into what it will take to migrate the Windows Thg packages to Qt5.
>
> Other than the dearth of PyQt5 binary packages, which I discussed earlier, the main blocker is now Kdiff3. There hasn’t been a Kdiff3 release in three years, and while they claim Qt5 support there have been no releases since they added the feature, and no-one else is making Windows binaries that I’ve found.
>
> Options:
>
> 1 - we un-bundle kdiff3 and tell people to download and install it separately, or some other merge tool
> 2 - someone builds and tests it against Qt-5.6 and sends me new binaries
> 3 - we package Kdiff3 statically linked with Qt4

Option 3 (or 2) is the best. Having reasonable differ/merger bundled
is a good thing for beginners and a reasonable default setting.

Michal

[Yuya Nishihara]

On Fri, 11 Aug 2017 19:08:08 +0200, Michał Sznajder wrote:
> > Other than the dearth of PyQt5 binary packages, which I discussed earlier, the main blocker is now Kdiff3. There hasn’t been a Kdiff3 release in three years, and while they claim Qt5 support there have been no releases since they added the feature, and no-one else is making Windows binaries that I’ve found.
> >
> > Options:
> >
> > 1 - we un-bundle kdiff3 and tell people to download and install it separately, or some other merge tool
> > 2 - someone builds and tests it against Qt-5.6 and sends me new binaries
> > 3 - we package Kdiff3 statically linked with Qt4
>
> Option 3 (or 2) is the best. Having reasonable differ/merger bundled
> is a good thing for beginners and a reasonable default setting.

I'm okay with 1. The bundled kdiff3 lacks CJK encoding support so I always
tell my coworkers to install kdiff3 separately.

> > Another problem I’ve encountered on Ubuntu Linux is that with PyQt5 the QSci widgets all have incorrect cursor positioning, it appears the cursor is positioned based on an integer font width but the font itself has a fractional (floating point) width. IE: when the cursor is at the end of 14 monospace characters, it is rendered between the 13th and 14th. It’s pretty disorienting. Can anyone else reproduce this? I’ve tried a few workarounds to resolve the problem, but was not successful.

Couldn't reproduce on Debian sid.

** System: Linux mimosa 4.11.0-2-amd64 #1 SMP Debian 4.11.11-1 (2017-07-22) x86_64
** Qt-5.7.1 PyQt-5.7 QScintilla-2.9.3