Dear all,
I would like to propose that we release our LyX 2.3.0 alpha binaries for
Mac and Windows with Qt 5.9.0beta1 and that we release our final LyX
2.3.0 binaries with the final release of Qt 5.9.0.
I think there is an advantage to releasing the final LyX version with
the same Qt x.y version (here I am considering Qt 5.9beta1 and Qt 5.9
final the "same") as our pre-releases. I make this claim for a few
reasons:
- In the past, sometimes when we switched Qt versions, this involved
switching versions of compilers, which caused complications (e.g. not
being able to compile) and worse, bugs (either a bug from the new
compiler or a bug in LyX that was exposed by the new computer and not
caught because not much testing).
- I consider Qt development snapshots to be quite stable. This is only
based on my own observations.
- (I'm not sure of this one) We are more likely to get a Qt bug that is
a regression fixed. Normally, regressions with respect to the most
recently released version are given higher priority than regressions
with respect to the version before the previous released version.
Now some arguments specific to the current situation of Qt development:
- From what I understand, there will be no Qt 5.8.1 released. Instead,
these improvements will only go in to Qt 5.9.0. See [1], in
particular, the following quote:
Unless there are significant security or other issues found,
we are not planning to provide any patch releases for Qt 5.8
to make sure that we release Qt 5.9.0 with the planned
schedule. Qt Support will assist customers to overcome
possible issues found in Qt 5.8.0 release.
- There will be Qt 5.9.z patch versions released [1]. This increases the
probability that our LyX 2.3.1 will be released with the same major Qt
version as 2.3.0, which minimizes regressions due to Qt.
- The alpha release met its deadline, and the beta release was only two
days late. Although historically Qt's releases have missed many of
their targets, there seems to be an increased effort by Qt's
development team to meet the deadlines on their release plan [2]. This
leads me to think that even if there are delays and Qt 5.9.0 final is
not released on 31 May, as currently planned, it will be released not
too late after.
Stephan and Uwe, do you forsee any problems with this plan? Would it be
a lot of extra work to also compile our alpha and beta with Qt 5.8.0,
but to not post these installers. This way, if we have a tester that
reports a regression that no LyX developer can reproduce, at least we
can ask them if they can reproduce with the LyX+5.8.0 installers (which
we would just send privately and not post publicly)? This could tell us
whether a regression is due to LyX or Qt.
If there is agreement on the above plan, I would then like to ask if
anyone could test LyX master with Qt 5.9beta going forward. There are no
offline installers for Qt 5.9beta, but there are binaries available
through the online installer. If you haven't used this yet, it is pretty
easy (at least in my experience on Linux). The most painful part for me
was answering questions before being able to download the online
installer.
I have only briefly tested Qt 5.9beta. At least, master compiles against
it without error. My plan would be to test master with 5.9beta, and for
work where I need a stable LyX, I would use 2.2.x branch compiled with
5.9beta.
Scott
[1]
http://blog.qt.io/blog/2017/02/22/qt-roadmap-for-2017/
[2]
https://wiki.qt.io/Qt_5.9_Release
signature.asc
Description: PGP signature
