a bit of log from #kde-devel - related to Xlib+Xdefs includes issues

irc.log (text/plain), 14.46 KB, created by Maciej Mrozowski on 2009-03-26 17:49:15 UTC

<reavertm> just to warn you, that Xdefs will probably will be added to Xrender include files soon and will need remobing some definitions from fixx11h.h (like Bool type being redefined)
<dfaure> fredrikh: I blame vandenoever
<pinda> Sho_: cool, nice to hear.
--> Kaboon (i=[e-mail]) has joined #kde-devel
<dashbot> 4.3/kdebindings r943952 failed (details under http://developer.kde.org/~dirk/dashboard/ ):
<dashbot> aseigo: smoke/plasma/x_1.cpp: In member function 'void x_Plasma__AbstractRunner::x_26(Smoke::StackItem*) const':
<dashbot> aseigo: smoke/plasma/x_1.cpp:808: error: 'RunnerSyntax' was not declared in this scope
<-> neoclust`aw is now known as neoclust
<pinda> Sho_: building now :)
<fredrikh> dfaure: and even if that's changed to simply list all of them, KFileMetaInfoItem::name() returns incredibly useful values, such as "http://freedesktop.org/standards/xesam/1.0/core#artist" :)
<dfaure> nice.
<fredrikh> and that's supposed to be a translated user friendly string... according to the api docs
<thiago> Alt+F2 doesn't do anything anymore
<thiago> no global shortcuts work
<dfaure> fredrikh: yep, that class was butchered with the strigi port, if you ask me
<pinda> Sho_: you think it will be ready for a release somewhere around kde 4.3?
<-- moofang (n=[e-mail]) has quit (Remote closed the connection)
<dfaure> reavertm: not all users of fixx11h.h use XRender....
<Sho_> pinda: Not sure about a non-alpha/beta final release, but I do hope/expect to have some form of public release out by then, yes
<pinda> Sho_: cool.
--> theseinfeld (n=me@88.193.141.67) has joined #kde-devel
<pinda> Sho_: anyway, build has finished, lets try it out :)
<-- pinda (n=rob@130.89.165.70) has quit (Remote closed the connection)
<Sho_> pinda: The biggest items left right now are KDE4-ifying the config dialog and some work on the chat view (bringing things like marker lines and nickname context menus back), other than that it's an assortment of polishing/bugfixing.
--> ddenis (n=denis@62.70.27.104) has joined #kde-devel
<pinotree> Sho_: commit that configdialog!!!!!! please :)
<Sho_> yes yes .. :)
<reavertm> dfaure https://bugs.gentoo.org/show_bug.cgi?id=263823 - we (gentoo) will need to workaround it somehow anyway
--> pinda (n=[e-mail]) has joined #kde-devel
<bugbot> bugs.gentoo.org bug 263823 in Gentoo Linux (Applications) "x11-proto/renderproto-0.9.3-r1 breaks compilation of KDE4 stuff" [Normal,New]
<Sho_> pinotree: you know me, I have the annoying tendency to do things like that out of sight and commit silly huge diffs at the end ;)
-*- pinda on konversation 4 :)
<Sho_> pinda: [18:15] <Sho_> pinda: The biggest items left right now are KDE4-ifying the config dialog and some work on the chat view (bringing things like marker lines and nickname context menus back), other than that it's an assortment of polishing/bugfixing.
--> rohyt (i=d2d43dfb@gateway/web/ajax/mibbit.com/x-af71b9f3af5c84af) has joined #kde-devel
<-> trueg is now known as trueg_away
--> bobesponja (n=bobespon@190.42.4.93) has joined #kde-devel
--> hoisn (n=[e-mail]) has joined #kde-devel
<-- DeusIX (n=[e-mail]) has quit (kubrick.freenode.net irc.freenode.net)
<-- rdieter (n=[e-mail]) has quit (kubrick.freenode.net irc.freenode.net)
<-- valente_ (n=[e-mail]) has quit (kubrick.freenode.net irc.freenode.net)
<-- SaroEngels (n=[e-mail]) has quit (kubrick.freenode.net irc.freenode.net)
<-- spstarr_work (n=spstarr@192.219.104.10) has quit (kubrick.freenode.net irc.freenode.net)
<-- gamaral|away (i=gamaral@kde/developer/gamaral) has quit (kubrick.freenode.net irc.freenode.net)
<-- DxOffEagle (n=[e-mail]) has quit (kubrick.freenode.net irc.freenode.net)
<-- keytoaster (n=tobias@gentoo/developer/keytoaster) has quit (kubrick.freenode.net irc.freenode.net)
<-- jpwhiting (n=jeremy@kde/developer/whiting) has quit (kubrick.freenode.net irc.freenode.net)
<-- trueg_away (n=[e-mail]) has quit (kubrick.freenode.net irc.freenode.net)
<-- das_netz (n=dasnetz@189.130.15.252) has quit (kubrick.freenode.net irc.freenode.net)
<-- tictric (n=Manfred@unaffiliated/tictric) has quit (kubrick.freenode.net irc.freenode.net)
<Sho_> gah, could Freenode please fix their network
--> spstarr_work (n=spstarr@192.219.104.10) has joined #kde-devel
--> gamaral|away (i=gamaral@kde/developer/gamaral) has joined #kde-devel
--> DeusIX (n=[e-mail]) has joined #kde-devel
--> keytoaster (n=tobias@gentoo/developer/keytoaster) has joined #kde-devel
--> DxOffEagle (n=[e-mail]) has joined #kde-devel
--> jpwhiting (n=jeremy@kde/developer/whiting) has joined #kde-devel
--> valente_ (n=[e-mail]) has joined #kde-devel
--> SaroEngels (n=[e-mail]) has joined #kde-devel
--> trueg_away (n=[e-mail]) has joined #kde-devel
--> rdieter (n=[e-mail]) has joined #kde-devel
--> das_netz (n=dasnetz@189.130.15.252) has joined #kde-devel
--> tictric (n=Manfred@unaffiliated/tictric) has joined #kde-devel
<dfaure> reavertm: so Bool isn't a #define anymore??
<-- alediaferia (n=alediafe@kde/developer/adiaferia) has quit ("No Ping reply in 30 seconds.")
<pinda> Sho_: seems to work good enough for ordinary use :)
--> alediaferia (n=[e-mail]) has joined #kde-devel
<-- smarter (n=smarter@ubuntu/member/smarter) has quit (Client Quit)
-*- dfaure agrees with pinda ;)
--> smarter (n=smarter@ubuntu/member/smarter) has joined #kde-devel
<thiago> plasma-desktop(22576)/plasma DesktopView::adjustSize: screen 1 geom QRect(1680,0 1680x1050)
<reavertm> dfaure: in Xdefs.h it's defined as int apparently (using typedef)
<Sho_> pinda: yes, it should. if you see any non-obvious bugs, or even a crash with a useful backtrace, run over to the bug tracker please :)
<thiago> if it gets the right size, why is it not drawing the right image?
<pinda> Sho_: I will
-*- thiago blames it on the same code that tries to divide by zero
<dfaure> reavertm: but fixx11h.h says #ifdef Bool around it all. So that code is only triggered if it's also a #define (probably from X11). So they are the ones defining things incompatibly, no?
-*- pinda enjoys being able to have nice plasma notifications when someone types my nick :)
<thiago>                 int height = width * imgSize.height() / imgSize.width();
<reavertm> dfaure: it is guarded by ifndef _XTYPEDEF_BOOL (it defines _XTYPEDEF_BOOL as well) check, so you maybe you could use it
<thiago> (gdb) p imgSize.width()
<thiago> $3 = 0
<-- emilsedgh (n=emilsedg@217.218.77.100) has quit (Read error: 104 (Connection reset by peer))
<thiago>             imgSize *= ratio;
<thiago> (gdb) p ratio
<thiago> $4 = -nan(0x400000)
<-- Seli (n=llunak@nat/suse/x-21421472e61ffae1) has quit (Remote closed the connection)
<thiago> right, if you multiply by -nan, you get weird results
<-> gamaral|away is now known as gamaral
<maelcum|konv> what do you expect when multiplying a number with an un-number? :P
<reavertm> dfaure: http://dpaste.com/19569/ - Xdefs from 1.5.3 xserver release
<thiago> I bet this is an uninitialised memory access
<dfaure> on that note, plasma crashed...
<thiago> I don't see the variable 'ratio' being initialised anywhere
<-> gamaral is now known as gamaral|away
<-- Morpheus (n=kvirc@85.124.170.171) has quit ("KVIrc Insomnia 4.0.0, revision: 3147, sources date: 20090224, built on: 2009/03/16 09:21:51 UTC 3147 http://www.kvirc.net/")
<thiago> plasma-desktop(22576)/plasma RenderThread::run: loaded with QSize(768, 1024) nan
<dfaure> reavertm: this is just the same as the Xdefs.h I have here
<-> mornfall_ is now known as mornfall
<SadEagle> thiago: wait, -NaN?
<thiago> SadEagle: yes
<thiago>             ratio = m_ratio;
<thiago> so why does m_ratio have a bad value?
<SadEagle> reavertm: if it's not a #define, it's not a problem
--> vorian__ (n=steve@168.158.129.93) has joined #kde-devel
<-- vorian (i=steve@freenode/staff/vorian) has quit (Nick collision from services.)
<-- vorian__ (n=steve@168.158.129.93) has quit (Client Quit)
<thiago> nothing in valgrind
--> vorian (i=steve@freenode/staff/vorian) has joined #kde-devel
<thiago> people should run their applications with QT_FATAL_WARNINGS turned on
<thiago> in fact, I should turn that on in KApplication, just to annoy people
<dfaure> plasma surely spits out a large amount of warnings indeed
<thiago> Breakpoint 2, RenderThread::setRatio (this=0x8e900b8, ratio=-nan(0x400000))
<reavertm> dfaure: I guess the point of adding Xdefs to Xrender includes was to make xrender apps buildable just using Xrender.h - I don't know whether it's policy of xorg guys to make those x11 extensions headers sufficient - anyway patch looks like - http://dpaste.com/19574/ - and will be proposed upstream
<-- Jucato (n=Jucato@kde/developer/jucato) has quit ("Caught sigterm, terminating...")
<-- marc_kdab (n=[e-mail]) has quit (Read error: 110 (Connection timed out))
<-- theseinfeld (n=me@unaffiliated/theseinfeld) has quit (Remote closed the connection)
<thiago>     m_ratio = boundingRect().width() / boundingRect().height();
<thiago> (gdb) p boundingRect()
<thiago> $9 = {xp = 49, yp = -12, w = 0, h = 0}
<thiago> 0 / 0 = NaN
<-- fosterfeld (n=[e-mail]) has quit
<dfaure> reavertm: you're still shooting in the dark here. Use gcc -E to find out what the fixx11h.h code really expands to. I still see no reason why it should fail.
<-> cylmor_ is now known as cylmor
<dfaure> if it really expands to typedef int XBool; typedef XBool Bool;     ... hmm does that clash with typedef int Bool ?
<dfaure> is gcc that stupid?
<SadEagle> It only does it if Bool is #define'd
<thiago> yes
<thiago> you can only typedef something once
<jpwhiting> Sho_: yakuake is yours, right?
<thiago> typedef int XBool; typedef int XBool; <--- wrong
<dfaure> ok
--> montoya (n=[e-mail]) has joined #kde-devel
<-- ervin (n=kdab@kde/ervin) has quit (Read error: 104 (Connection reset by peer))
--> gregarei_ (n=[e-mail]) has joined #kde-devel
<reavertm> hence just add check #ifndef _TYPEDEF_BOOL in fixx11h.h and you should be fine I guess
<dfaure> no
--> jmbsvicetto (i=jmbsvice@gentoo/developer/jmbsvicetto) has joined #kde-devel
<dfaure> reavertm: just make that patch correct and include <Xlib.h> first?
<Sho_> jpwhiting: yes
<dfaure> reavertm: then Bool will be #defined to int, and Xdefs won't do anything
<jpwhiting> Sho_: just built from trunk with trunk konsole, and get a crash when I try to change the height, do you get that too?
<Sho_> jpwhiting: no, I'm not on trunk though
<reavertm> dfaure: I don't follow - make what patch correct? renderproto patch or ?
<jpwhiting> the backtrace is not very useful, but I also got a crash before when I was using arch 4.2 packages of yakuake, and kde
<dfaure> reavertm: yes
<jpwhiting> but no backtrace then because it had no symbols =)
<Sho_> jpwhiting: backtrace, and when it points to yakuake code file it for yakuake; when it points to konsole code, file it for konsole :)
<-- mjansen (n=[e-mail]) has quit (Read error: 110 (Connection timed out))
<jpwhiting> Sho_: ok, will do
<reavertm> dfaure: from X11 and our probably as well, point of view, every x11 extension header should compile (should be self-sufficient - hence shoule include Xdefs.h if needed)
<reavertm> how would you propose to fix this patch?
<dfaure> reavertm: it seems to me that Xdefs.h itself isn't self-sufficient, it defines things that conflict with Xlib.h. I would propose s/Xdefs/Xlib/ in that patch
<Sho_> jpwhiting: (actually while I use 4.2.1, my Konsole is a from trunk since I recently made some fixes there, but it's already two weeks old again)
<jpwhiting> http://www.pastebin.ca/1373138 <-- here's the backtrace, looks like qt
<jpwhiting> but not backtrace in thread 0...
<Sho_> hm, actually morel ike one week
<jpwhiting> *no
<Sho_> *more like
<jpwhiting> Sho_: I'm wondering if it maybe built against my distro's 4.5 qt, and links to my qt-copy
<Sho_> hm maybe
<Sho_> I have a system 4.5 here
<Sho_> no qt-copy
<reavertm> are you sure about those confllicts? build log shows Xdefs.h:49 and fixx11h.h:94 are conflicting
<reavertm> dfaure: ^
<dfaure> sigh
<dfaure> just read the files
--> blaze (n=blaze@unaffiliated/blaze) has joined #kde-devel
<dfaure> one says typedef and one says #define. that's a conceptual conflict.
<dfaure> but since in your case Xlib.h isn't included, obviously it's not in the picture
<dfaure> however it would be included, since that's what everyone else includes
<SadEagle> fixx11h.h only messed with Bool if it's #define'd.
<dfaure> so everyone else gets #defines, but XRender would get the typedef?
<-- blaze (n=blaze@unaffiliated/blaze) has quit (Client Quit)
<dfaure> SadEagle: yes it is
--> smarter_ (n=smarter@ubuntu/member/smarter) has joined #kde-devel
<SadEagle> Xdefs.h only messes with it if it's NOT defined
<SadEagle> Seems to me something other than KDE is trying to both define and typedef it.
<andris> sounds like a real mess :
<dfaure> SadEagle: and what happens if you include Xdefs.h and then Xlib.h?  you get the typedef AND the define?
<andris> :)
<SadEagle> dfaure: probably.
<reavertm> then proper fix would be to remove those #defines from Xlib and include Xdefs there
<dfaure> SadEagle: that's exactly what breaks fixx11h.h
<SadEagle> So checking for _XTYPEDEF_BOOL is probably reasonable.
--> t-nor (n=[e-mail]) has joined #kde-devel
<-- smarter (n=smarter@ubuntu/member/smarter) has quit (No route to host)
<SadEagle> reavertm: no, that's not proper.
<SadEagle> That changes behavior of Xlib.h
<dfaure> <irony> sure let's use even more implementation details of X11 that are going to change at some point...
<SadEagle> Well, we can't rely on a fixed X11 version anyway
<dfaure> but yeah that seems like the only solution, if Xdefs+Xlib can lead to this, independently from that Xrender.h patch.
--> toma (n=toma@kde/toma) has joined #kde-devel
<reavertm> I guess we need to push them further with headers/definitions cleanup...
<dfaure> well afaics they have cleaned up. They did the "right thing" in newer headers (Xdefs) and had to keep the old crap (Xlib) :/
<SadEagle> Depends on what you mean by 'we'.
<-- Kaboon (i=[e-mail]) has quit
<dfaure> (...which makes the whole situation even worse)
<SadEagle> From KDE PoV, any changes will either create new problems, or make no difference