We CAN'T break up the installer. Some things work only on MSVC (e.g.: WMI backend for Solid) and some only on MinGW. It's a matter of personal taste, period.<div>Releases for different compilers being independent is ok, but... don't we already do that? At least we did it when we started supporting mingw on the installer. As long as there is no big gap (2 weeks?) on releases for each compiler sounds okay. Most of us use only one compiler. And most things that are fixed for one compiler is also fixed for the other.</div>
<div><br></div><div><br><div class="gmail_quote">2010/11/6 Thomas Friedrichsmeier <span dir="ltr"><<a href="mailto:thomas.friedrichsmeier@ruhr-uni-bochum.de">thomas.friedrichsmeier@ruhr-uni-bochum.de</a>></span><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><div class="im">On Saturday 06 November 2010, Casper van Donderen wrote:<br>
> Everybody should take his compiler of choice<br>
<br>
</div>Yes, and that's exactly the philosophy I'm trying to enable with respect to<br>
creating releases. Everybody take the compiler they care about, and only that<br>
one, instead of making releases artificially cumbersome by requiring to produce<br>
a complete set binaries for the complete set of compilers.<br>
<div class="im"><br>
> and maybe make the default<br>
> the platform default:<br>
<br>
</div>To me, *personally*, GCC 32bit is the one hard requirement in the compiler<br>
department, because an external dependency of my application supports only<br>
MinGW on windows. And so that's why I'm willing to help with MinGW, and that's<br>
why I *personally* don't care about any other compiler.<br>
<br>
But you're right in that it's a dead-end to try to discuss one compiler<br>
against another, and I see we may be getting hung up on the side issue of<br>
which should be the default.<br>
<br>
So I'll modify my proposal for yet another short-term strategy (see also [1])<br>
to allow breaking up the release process into smaller portions:<br>
- Multi-compiler support will remain in the next version of the installer, but<br>
instead of working on point releases, the default choice of releases will be<br>
trimmed down to:<br>
* stable-latest<br>
* unstable-latest<br>
* nightly-latest<br>
- These directories will differ from the current layout in that stable-latest<br>
may contain different versions of packages for different compilers. Perhaps at<br>
times MinGW will be a couple versions ahead, and perhaps sometimes MSVC will<br>
be some versions ahead. So once again, releases for the different compilers can<br>
be made independently of each other.<br>
- As long as there are no binary compatibility breakages (as coming up with<br>
KDE 4.6, AFAIK), this concept will even allow to create releases in a more<br>
incremental fashion, e.g. uploading an updated kdelibs, but leaving kdegames<br>
at its old version until a volunteer picks up that one.<br>
- Point releases will merely be archive snapshots, and will be completely<br>
hidden from the user (and perhaps even from the mirrors). Joe User only cares<br>
about the latest available stable/unstable/nightly releases, anyway.<br>
<br>
I think it's still a good idea to break up the installer into one incarnation<br>
for each compiler, and to decide on a default. But I'll leave that for another<br>
day.<br>
<br>
Regards<br>
Thomas<br>
<br>
[1] <a href="http://lists.kde.org/?l=kde-windows&m=128689460909832&w=2" target="_blank">http://lists.kde.org/?l=kde-windows&m=128689460909832&w=2</a><br>
<br>_______________________________________________<br>
Kde-windows mailing list<br>
<a href="mailto:Kde-windows@kde.org">Kde-windows@kde.org</a><br>
<a href="https://mail.kde.org/mailman/listinfo/kde-windows" target="_blank">https://mail.kde.org/mailman/listinfo/kde-windows</a><br>
<br></blockquote></div><br></div>