Re: split_marker crash
On Thu, 22 Nov 2018 at 10:21, Ben Gamari wrote: > Steven Shaw writes: > > > Curious if the GHC libraries _are_ still built with the `-split-objs` > flag, > > as pointed out in the latest documentation. > > > Good catch; the build system uses split sections by default if > supported. If not it instead uses split objects, if supported. I do not > believe there are any cases where we support split objects but not split > sections. Consequently I believe we should always use split sections. > 👍 split sections sounds best. ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
Steven Shaw writes: > Curious if the GHC libraries _are_ still built with the `-split-objs` flag, > as pointed out in the latest documentation. > Good catch; the build system uses split sections by default if supported. If not it instead uses split objects, if supported. I do not believe there are any cases where we support split objects but not split sections. Consequently I believe we should always use split sections. Cheers, - Ben signature.asc Description: PGP signature ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
Curious if the GHC libraries _are_ still built with the `-split-objs` flag, as pointed out in the latest documentation. https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/phases.html?highlight=split-objs#ghc-flag--split-objs ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
Phyx writes: > On Mon, Nov 19, 2018 at 8:33 PM Ben Gamari wrote: > >> Andreas Klebinger writes: >> >> > Hello, >> > >> > I'm fine with reverting for now. But could you give me a way to >> > reproduce this error? >> > >> > I've not seen crashes on either linux and windows in various configs. >> > >> I suspect that Simon is building with SplitObjects enabled. To be >> honest, I would really like to remove this feature; SplitSections is >> better in nearly every regard. However, we have been stalled since >> SplitSections doesn't quite work yet on Windows (or, IIRC, the toolchain >> is prohibitively slow when it's used). I believe Tamar was working on >> fixing this. >> > > SplitObjects has been off in favor of SplitSections on Windows since Aug. > https://github.com/ghc/ghc/commit/23774c98f1368b41515cbd5223b87ea6dbf644e1 > Ahh, I had forgotten about this. I'll need to check this when I'm back home but I suspect this means that we can finally rip out (or at least deprecate) split object support for 8.8. Cheers, - Ben signature.asc Description: PGP signature ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
RE: split_marker crash
Simon Peyton Jones via ghc-devs writes: > I do indeed have SplitObs=YES in my validate.mk. > Ahh, so perhaps Andreas's finding isn't the culprit afterall. Looking at the source it looks like `split_marker_entry` is indeed emitted for split objects support.c > In the past at least, that made binaries quite a bit smaller. > Happy to change it to SplitSections=YES if that is more kosher these days? > We generally recommend SplitSections at this point. Not only is it simpler, but it may reduce build times, result in smaller binaries, and generally stays closer to what C compilers do, avoiding toolchain bugs. Cheers, - Ben signature.asc Description: PGP signature ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
On Mon, Nov 19, 2018 at 8:53 PM Simon Peyton Jones wrote: > I do indeed have SplitObs=YES in my validate.mk. > > In the past at least, that made binaries quite a bit smaller. > > Happy to change it to SplitSections=YES if that is more kosher these days? > Yes, SplitObjs is no longer supported, will also cause an extremely slow compilation depending on your optimization level and ways. Removing the line should let the compiler default to the best choice, in this case it should go to SplitSections on its own when compiling stage2. Regards, Tamar > > Simon > > > > *From:* Andreas Klebinger > *Sent:* 19 November 2018 20:38 > *To:* Ben Gamari > *Cc:* Phyx ; Simon Peyton Jones < > simo...@microsoft.com>; ghc-devs@haskell.org > *Subject:* Re: split_marker crash > > > > I might already have found the specific cause. > > It seems with split sections we generate a dummy CmmProc, which has as > entry label > > error (split_sections_entry) > > My patch likely introduces strictness on that field where it was lazy > before. > If this is the cause I expect to have a patch up in a hour or two and will > merge > it after it validates. > > Cheers, > Andreas > > Ben Gamari schrieb: > > Andreas Klebinger > writes: > > > > Hello, > > > > I'm fine with reverting for now. But could you give me a way to > > reproduce this error? > > > > I've not seen crashes on either linux and windows in various configs. > > > > I suspect that Simon is building with SplitObjects enabled. To be > > honest, I would really like to remove this feature; SplitSections is > > better in nearly every regard. However, we have been stalled since > > SplitSections doesn't quite work yet on Windows (or, IIRC, the toolchain > > is prohibitively slow when it's used). I believe Tamar was working on > > fixing this. > > > > Cheers, > > > > - Ben > > > ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
On Mon, Nov 19, 2018 at 8:33 PM Ben Gamari wrote: > Andreas Klebinger writes: > > > Hello, > > > > I'm fine with reverting for now. But could you give me a way to > > reproduce this error? > > > > I've not seen crashes on either linux and windows in various configs. > > > I suspect that Simon is building with SplitObjects enabled. To be > honest, I would really like to remove this feature; SplitSections is > better in nearly every regard. However, we have been stalled since > SplitSections doesn't quite work yet on Windows (or, IIRC, the toolchain > is prohibitively slow when it's used). I believe Tamar was working on > fixing this. > SplitObjects has been off in favor of SplitSections on Windows since Aug. https://github.com/ghc/ghc/commit/23774c98f1368b41515cbd5223b87ea6dbf644e1 SplitObjects is not usable on Windows anymore so it was disabled https://ghc.haskell.org/trac/ghc/ticket/15051 > Cheers, > > - Ben > ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
RE: split_marker crash
I do indeed have SplitObs=YES in my validate.mk. In the past at least, that made binaries quite a bit smaller. Happy to change it to SplitSections=YES if that is more kosher these days? Simon From: Andreas Klebinger Sent: 19 November 2018 20:38 To: Ben Gamari Cc: Phyx ; Simon Peyton Jones ; ghc-devs@haskell.org Subject: Re: split_marker crash I might already have found the specific cause. It seems with split sections we generate a dummy CmmProc, which has as entry label > error (split_sections_entry) My patch likely introduces strictness on that field where it was lazy before. If this is the cause I expect to have a patch up in a hour or two and will merge it after it validates. Cheers, Andreas Ben Gamari schrieb: Andreas Klebinger <mailto:klebinger.andr...@gmx.at> writes: Hello, I'm fine with reverting for now. But could you give me a way to reproduce this error? I've not seen crashes on either linux and windows in various configs. I suspect that Simon is building with SplitObjects enabled. To be honest, I would really like to remove this feature; SplitSections is better in nearly every regard. However, we have been stalled since SplitSections doesn't quite work yet on Windows (or, IIRC, the toolchain is prohibitively slow when it's used). I believe Tamar was working on fixing this. Cheers, - Ben ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
Andreas Klebinger writes: > I might already have found the specific cause. > > It seems with split sections we generate a dummy CmmProc, which has as > entry label > > error (split_sections_entry) > > My patch likely introduces strictness on that field where it was lazy > before. If this is the cause I expect to have a patch up in a hour or > two and will merge it after it validates. > Ahhh, yes, that sounds like it may be the culprit. Good sleuthing! Cheers, - Ben signature.asc Description: PGP signature ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
I might already have found the specific cause. It seems with split sections we generate a dummy CmmProc, which has as entry label > error (split_sections_entry) My patch likely introduces strictness on that field where it was lazy before. If this is the cause I expect to have a patch up in a hour or two and will merge it after it validates. Cheers, Andreas Ben Gamari schrieb: Andreas Klebinger writes: Hello, I'm fine with reverting for now. But could you give me a way to reproduce this error? I've not seen crashes on either linux and windows in various configs. I suspect that Simon is building with SplitObjects enabled. To be honest, I would really like to remove this feature; SplitSections is better in nearly every regard. However, we have been stalled since SplitSections doesn't quite work yet on Windows (or, IIRC, the toolchain is prohibitively slow when it's used). I believe Tamar was working on fixing this. Cheers, - Ben ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
Andreas Klebinger writes: > Hello, > > I'm fine with reverting for now. But could you give me a way to > reproduce this error? > > I've not seen crashes on either linux and windows in various configs. > I suspect that Simon is building with SplitObjects enabled. To be honest, I would really like to remove this feature; SplitSections is better in nearly every regard. However, we have been stalled since SplitSections doesn't quite work yet on Windows (or, IIRC, the toolchain is prohibitively slow when it's used). I believe Tamar was working on fixing this. Cheers, - Ben signature.asc Description: PGP signature ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
Re: split_marker crash
Hello, I'm fine with reverting for now. But could you give me a way to reproduce this error? I've not seen crashes on either linux and windows in various configs. Cheers, Andreas Ben Gamari schrieb: Simon Peyton Jones via ghc-devs writes: OK I have verified that * This split_marker crash happens on a clean HEAD build * Reverting "NCG: New code layout algorithm.", 575515b4909f3d77b3d31f2f6c22d14a92d8b8e0, solves the problem. Andreas: I propose to revert in HEAD unless you have a rapid fix, or believe that is somehow my fault. (Or maybe someone else can revert.) Simon, are you using split-objs by any chance? Regardless, yes, let's revert until we work out what is going on here. Cheers, - Ben ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
RE: split_marker crash
Simon Peyton Jones via ghc-devs writes: > OK I have verified that > > * This split_marker crash happens on a clean HEAD build > * Reverting "NCG: New code layout algorithm.", > 575515b4909f3d77b3d31f2f6c22d14a92d8b8e0, solves the problem. > Andreas: I propose to revert in HEAD unless you have a rapid fix, or believe > that is somehow my fault. > (Or maybe someone else can revert.) Simon, are you using split-objs by any chance? Regardless, yes, let's revert until we work out what is going on here. Cheers, - Ben signature.asc Description: PGP signature ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
RE: split_marker crash
OK I have verified that * This split_marker crash happens on a clean HEAD build * Reverting "NCG: New code layout algorithm.", 575515b4909f3d77b3d31f2f6c22d14a92d8b8e0, solves the problem. Andreas: I propose to revert in HEAD unless you have a rapid fix, or believe that is somehow my fault. (Or maybe someone else can revert.) Simon From: Simon Peyton Jones Sent: 19 November 2018 14:53 To: 'ghc-devs@haskell.org' Subject: split_marker crash Urrgh. I have this panic in a clean validate (sh validate -fast) on my branch wip/T15809. ghc-stage1: panic! (the 'impossible' happened) (GHC version 8.7.20181119 for x86_64-unknown-linux): split_marker_entry Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug libraries/ghc-prim/ghc.mk:4: recipe for target 'libraries/ghc-prim/dist-install/build/GHC/Types.o' failed Yes, it's a branch not identical to HEAD, but I have done nothing concerning split_markers. Sigh. What should I do? Simon ___ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs