[Bug 52420] New: Wine 7.0 Release Compiled with Clang on Arch Produces a Broken Build
https://bugs.winehq.org/show_bug.cgi?id=52420 Bug ID: 52420 Summary: Wine 7.0 Release Compiled with Clang on Arch Produces a Broken Build Product: Wine Version: 7.0 Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: critical Priority: P2 Component: -unknown Assignee: wine-bugs(a)winehq.org Reporter: nekoNexus(a)protonmail.ch Distribution: --- Created attachment 71674 --> https://bugs.winehq.org/attachment.cgi?id=71674 Arch PKGBUILD File for Wine 7.0 Release Using the default Arch makepkg flags (from /etc/makepkg.conf - by default this uses GCC so the user either has to change this to Clang or export to Clang in the PKGBUILD), using this example PKGBUILD produces a WoW64 build but throws stack overflow errors upon calling winecfg and doesn't run winecfg. The same is true if all of the miscellaneous variables exported along with Clang are not set; this will still occur. Un-setting lld and/or any of the llvm variables bakes no difference. This topic also has a page an the Arch forums: https://bbs.archlinux.org/viewtopic.php?id=273313 -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 nekoNexus(a)protonmail.ch changed: What |Removed |Added ---------------------------------------------------------------------------- Distribution|--- |ArchLinux -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 Fabian Maurer <dark.shadow4(a)web.de> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |dark.shadow4(a)web.de --- Comment #1 from Fabian Maurer <dark.shadow4(a)web.de> --- I usually use the following configure:
PKG_CONFIG_PATH=/usr/lib32/pkgconfig CC="clang" CFLAGS="-O3" ../wine-git/configure --without-cups --without-mingw
This works for me when compiling 32bit wine from source. Can you try if that works for you as well? Or what you make different? -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 --- Comment #2 from nekoNexus(a)protonmail.ch --- Over here I go over exactly what was different; this other issue is for GCC specifically but the same PKGBUILD is used and I explain that the variables are all reset in the PKGBUILD and the only things that seem to be different are the CFLAGS: https://bugs.winehq.org/show_bug.cgi?id=52419#c7 -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 --- Comment #3 from nekoNexus(a)protonmail.ch --- For that other issue, the -fno-plt flag was the problem; however, removing it doesn't fix compiling with Clang. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 --- Comment #4 from nekoNexus(a)protonmail.ch --- To clarify again, it *does* compile, it just doesn't run -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 nekoNexus(a)protonmail.ch changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #71674|0 |1 is obsolete| | --- Comment #5 from nekoNexus(a)protonmail.ch --- Created attachment 71681 --> https://bugs.winehq.org/attachment.cgi?id=71681 Arch PKGBUILD File for Wine 7.0 Release I managed to get the stack overflow errors to go away but it still doesn't run clang builds. :/ -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 nekoNexus(a)protonmail.ch changed: What |Removed |Added ---------------------------------------------------------------------------- Summary|Wine 7.0 Release Compiled |Wine 7.0 WoW64 Release |with Clang on Arch Produces |Compiled with Clang on Arch |a Broken Build |Produces a Broken Build -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 nekoNexus(a)protonmail.ch changed: What |Removed |Added ---------------------------------------------------------------------------- Keywords| |source -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 nekoNexus(a)protonmail.ch changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #71681|0 |1 is obsolete| | --- Comment #6 from nekoNexus(a)protonmail.ch --- Created attachment 71693 --> https://bugs.winehq.org/attachment.cgi?id=71693 Arch PKGBUILD File for Wine 7.0 Release So, I managed to narrow it down a bit but not completely (I'm also updating the attachment with my findings for Arch makepkg repoduction): - If the clang build is pure clang, it will compile, but won't run (tested by running winecfg) - If the the clang build has mingw enabled, it will compile, but only if ld is used and lld isn't used - If mingw is enabled with a clang build, relro can't be used because mingw's ld doesn't like the "-z" argument for whatever reason - If the build succeeds with Clang and mingw (which it always will, provided the above two points are followed), the resulting WoW64 build will run but will throw one stack overflow error - If certain CFLAGS are passed to clang, the resulting build will throw 3 stack overflow errors; an example would be something like this:
export CFLAGS="-O3 -pipe -fno-plt -minline-all-stringops -fexceptions -Wall \ -Wformat -Werror=format-security \ -fstack-clash-protection -fcf-protection"
----- That said, as a reminder, even with the stack overflow errors Wine will still run but a few things are made clear: - Pure clang builds have some kind of problem preventing use, even with simple flag setups like: "-O3 -pipe -std=gnu89" - Clang + mingw builds run but I wouldn't trust a build spontaneously throwing overflow errors and - Clang + mingw builds require compromising relro -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 --- Comment #7 from nekoNexus(a)protonmail.ch ---
- If certain CFLAGS are passed to clang, the resulting build will throw 3 stack overflow errors; an example would be something like this:
export CFLAGS="-O3 -pipe -fno-plt -minline-all-stringops -fexceptions -Wall \ -Wformat -Werror=format-security \ -fstack-clash-protection -fcf-protection"
Sorry, I actually meant:
export CFLAGS="-O3 -pipe -fexceptions -Wall \ -Wformat -Werror=format-security \ -fstack-clash-protection -fcf-protection"
-- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=52420 --- Comment #8 from Ken Sharp <imwellcushtymelike(a)gmail.com> --- Does this still occur in the latest version of Wine? -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=52420 Ken Sharp <imwellcushtymelike@gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Keywords| |Abandoned? Status|UNCONFIRMED |NEEDINFO Ever confirmed|0 |1 Severity|critical |normal -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=52420 --- Comment #9 from Neko-san <nekoNexus@protonmail.ch> --- I just noticed I was asked for an update on this issue; I don't use Arch anymore (I use Gentoo now) but I'll have to get back to this when I have a moment soon. I had been physically separated from my PC for a while and I had been physically reunited with it but it's in need of Gentoo-specific maintenance that I need to address first before I can test whether Wine+Clang actually behaves now. Since I initially reported this bug, llvm-mingw also apparently either became a thing or "more recently" became a package (on Gentoo), and I don't think the Gentoo ebuild has any clear instructions on using that for "full" LLVM yet either. Back when I first wrote this report, it was just a Clang build with no mingw involvement (which technically is the target testing case but I'm pretty sure GCC/Clang + Mingw is more expected). -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
participants (2)
-
WineHQ Bugzilla -
WineHQ Bugzilla