Re: NEW: x11/qt6/{qtgrpc,qthttpserver,qtquick3dphysics,qtquickeffectmaker}

On Mon, May 01, 2023 at 01:37:46PM +0200, Rafael Sadowski wrote:
> On Mon May 01, 2023 at 09:16:12AM +0000, Klemens Nanni wrote:
> > On Mon, Apr 10, 2023 at 08:30:03AM +0200, Rafael Sadowski wrote:
> > > Hi ports hackers,
> > >
> > > here are some new Qt6 modules. OK to import?
> >
> > All build/package with complete WANTLIB, Makefiles look good.
>
> Thanks!
>
> >
> > I can't `cd $(make show=WRKSRC)' due to o=0 permissions.
> > Do we need qt(6?)-wide FIX_EXTRACT_PERMISSIONS=Yes?

x11/qt6/qt5compat and x11/qt5/qtvirtualkeyboard as random samples have the
same has the same issue:

$ stat -f%Sp `mk show=WRKSRC`
drwx------


I'd use FIX_EXTRACT_PERMISSIONS for them all, diff below.
Feedback? Objection? OK?

>
> I do not build with PORTS_PRIVSEP=yes.

Please do, otherwise ports may work for you (doing network, writing
outside of pobj, etc.), but fail for other porters and on bulk machines.

> > All but qtgrpc need update-plist to remove common directories.
>
> Hm, I have no changes here.

Do you have pkglocatedb package installed? update-plist(1) -f is the
default and checks against already registered files.

> > I see you define PX_OPENBSD in various places, but also define PX_LINUX, why?
>
> PX_LINUX is default and enabled and we use it but for some cases we need
> PX_OPENBSD.

Shouldn't all PX_LINUX checks eventually honour PX_OPENBSD as well?
Or do you want to piggy back PX_LINUX and only maintain the bare minimum of
OpenBSD specific macro foo?

> > This port uses syscall(2), i.e. `syscall(getthrid)', which sohuld be direct.
>
> "which sohuld be direct"? Why?

*NOT*, sorry

Syscalls must come from libc, lots of ports have been patched and fixed
upstream to not use direct syscall(2) but instead call into libc functions.

> > What's the story with s/NULL/0/ in the patches?
>
> This fix (our?) clang build. C++ template magic.

Can you mention this in a comment? It saves time and effort for anyone not
intimitely familiar with the C++ ecosystem; I can't tell if this is a legit
fix (in our OpenBSD ports case) or just a workaround.



Index: x11/qt5/Makefile.inc
===================================================================
RCS file: /cvs/ports/x11/qt5/Makefile.inc,v
retrieving revision 1.20
diff -u -p -r1.20 Makefile.inc
--- x11/qt5/Makefile.inc 13 Jul 2022 15:48:56 -0000 1.20
+++ x11/qt5/Makefile.inc 3 May 2023 17:15:20 -0000
@@ -1,5 +1,7 @@
ONLY_FOR_ARCHS ?= ${GCC4_ARCHS} ${CLANG_ARCHS}

+FIX_EXTRACT_PERMISSIONS ?= Yes
+
.include "Makefile.version"

# DIST_VERSION should be defined, e.g., when patch distfile gets issued,
Index: x11/qt6/Makefile.inc
===================================================================
RCS file: /cvs/ports/x11/qt6/Makefile.inc,v
retrieving revision 1.5
diff -u -p -r1.5 Makefile.inc
--- x11/qt6/Makefile.inc 28 May 2022 06:20:04 -0000 1.5
+++ x11/qt6/Makefile.inc 3 May 2023 17:15:27 -0000
@@ -1,5 +1,7 @@
ONLY_FOR_ARCHS ?= ${GCC4_ARCHS} ${CLANG_ARCHS}

+FIX_EXTRACT_PERMISSIONS ?= Yes
+
.include "Makefile.version"

VERSION ?= ${QT6_VERSION}