Date: Sun, 15 Jun 2008 16:51:18 +0200 From: Jan Henrik Sylvester <me@janh.de> To: emulation list freebsd <freebsd-emulation@freebsd.org> Cc: Gerald Pfeifer <gerald@FreeBSD.org> Subject: wine-kthread Message-ID: <48552C66.3030804@janh.de>
next in thread | raw e-mail | index | archive | help
I have got two questions about wine-kthread, one about the future of that patch and one about a current issue I got. 1. The need for wine-kthread: wine-kthread was removed from the port with 0.9.45 as the pthread version was supposed to "be better by now". For me, the popular game Warcraft 3 would not work without wine-kthread and it still segfaults immediately with wine(-pthread) as of 1.0.r4. (The German bsdforen wiki still recommends wine-kthread for quite a few games. I do not know if that is outdated.) In a long email exchange, Tijl Coosemans was extremely helpful analyzing the problem and explaining to me the underlying issues. (Thanks again!) The result was the patch-kthread to be found on the wiki. The patch actually changes two things: (a) Reintroduce wine-kthread as it seems to be needed. (b) Move wine from 0x7bf00400 to 0x81000400 as dri needs more heap for Warcraft 3 on certain occasions than is available with the standard base address. (I could reproducibly crash it with certain larger maps.) Tijl told me that at least the second issue would be fixed with FreeBSD 7, but my later testing at the time of 7.0-BETA4 indicated that this was not the case. Maybe it was just my failure to separate the correct part of the patch. Tijl did post on wine-devel about the problem: http://www.winehq.org/pipermail/wine-devel/2007-September/059426.html The only reply does not sound promising to me with my limited understanding of the underlying issue. Since it would probably break other application, Tijl wold me the the patch could not go back to the port. (See below: I found the first one myself that is affected.) Since FreeBSD 7 might have been able to help, the issue was postponed. 2. My current issue: As I tried to fix a problem with an old application I wrote with ancient Borland C++Builder 4, I tried to make BCB4 run using wine. BCB4 fails linking a project as long as wine is compiled with patch-kthread -- no matter if I start it with wine-ktread or wine(-pthread). The problem might be the 0x8 address: "Cannot reserve virtual memory at addr 90000000 for 33554432 bytes" -- I could provide more details, but I guess they are not needed. For the same project (a component needed for my application), linking works without patch-kthread. 3. The future of wine-kthread: (a) Would it be possible to build a wine version with wine-kthread at 0x81 and wine(-pthread) at 0x7b? (b) Is there a future for wine-kthread in the FreeBSD port, maybe as an OPTION? (c) Should the heap issue be solvable on FreeBSD 7? With (a), my current problem would be solved for the moment. I tried to analyze which files are affected by patch-kthread to be able to switch around and make building both versions easier, but since there seem to be quite a few files in the wine package that change with every recompile, I am not sure if I found the correct ones. My guess is that nothing outside loader/ is affected during build and only bin/wine, bin/wine-pthread, and bin/wine-kthread are affected for the package. Currently, (b) would only help me a little, since I use two more wine patches solving issues with Warcraft 3 network (same for others on Linux) and opengl (probably specific to my graphic chipset), but I hope these regressions will eventually get fixed in wine. (I did test the kthread issues to exist without further patches, too.) Anyhow, since the wine port recently got OPTIONS... Of course, (c) is the major point. It would be so much nicer not to have patches anymore, especially since they could become incompatible at any time. I repeated most of my earlier tests on wine-1.0.r4 with same results. Looking at the changes, 1.0.r5 should not change anything, but I could try again. I depend neither on Warcraft 3 nor on BCB4, thus the issue is not extremely high priority for me. Anyhow, I am interested in fixing it, but unfortunately my understanding of the issue is too limited to really try to fix it myself. Of course, I can provide more details, but this email already got too long and maybe they are not needed. Thanks in advance for any help, Jan Henrik
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?48552C66.3030804>