On Aug 26 12:27, Matt Smith wrote:>On Aug 26 13:10, Slawa Olhovchenkov wrote: >>>>Hardware error or memory exhausted >>> >>>It does appear to be something along those lines. It's not memory >>>exhaustion as it has around 2GB free at the point it fails. However I've >>>deleted /usr/obj and started the buildworld again and it failed in a >>>different place with the same error. I'm trying it again without -j4 to >>>see what happens. But isn't looking too good. :( >> >>Look like hardware error. >>RAM/CPU/MB > >Interestingly it *always* manages to succesfully compile clang etc and >it has no issues compiling things from ports. It fails compiling >something from lib like openssl or kerberos. Doesn't buildworld build >a bootstrap version of clang and then use that version to compile the >rest of it? I might try downgrading my sources back to the version >that I last succesfully compiled just to prove it one way or the other >to myself. >So, been doing some testing. It looks like a -j4 problem with the latest sources. If I buildworld with -j1 then it compiles with no issues at all. If I compile r286908 with -j4 then it compiles with no issues at all. If I try and compile r287155 with -j4 then I get the bus errors. So I'm not convinced at all that it's hardware related at the moment. -- Matt
On Wed, Aug 26, 2015 at 06:38:46PM +0100, Matt Smith wrote:> ... > So, been doing some testing. It looks like a -j4 problem with the latest > sources. If I buildworld with -j1 then it compiles with no issues at > all. If I compile r286908 with -j4 then it compiles with no issues at > all. If I try and compile r287155 with -j4 then I get the bus errors. So > I'm not convinced at all that it's hardware related at the moment. > ...While it is not a complete list of the updates I do to FreeBSD machines, I believe that <http://www.catwhisker.org/~david/FreeBSD/history/> has links to a selection that is reasonably representative of those I have been doing recently. For my builds, I use a -j value that is twice the reported value of `sysctl -n hw.ncpu`. For both my laptop and my build machine, that works out to -j16. For my work desktop (not depicted on the page), it's -j24. I'm not seeing the issues you're reporting. Peace, david -- David H. Wolfskill david at catwhisker.org Those who would murder in the name of God or prophet are blasphemous cowards. See http://www.catwhisker.org/~david/publickey.gpg for my public key. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 949 bytes Desc: not available URL: <http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20150826/e5f5e853/attachment.bin>
On Wed, 26 Aug 2015 18:38:46 +0100 Matt Smith <fbsd at xtaz.co.uk> wrote> On Aug 26 12:27, Matt Smith wrote: > >On Aug 26 13:10, Slawa Olhovchenkov wrote: > >>>>Hardware error or memory exhausted > >>> > >>>It does appear to be something along those lines. It's not memory > >>>exhaustion as it has around 2GB free at the point it fails. However I've > >>>deleted /usr/obj and started the buildworld again and it failed in a > >>>different place with the same error. I'm trying it again without -j4 to > >>>see what happens. But isn't looking too good. :( > >> > >>Look like hardware error. > >>RAM/CPU/MB > > > >Interestingly it *always* manages to succesfully compile clang etc and > >it has no issues compiling things from ports. It fails compiling > >something from lib like openssl or kerberos. Doesn't buildworld build > >a bootstrap version of clang and then use that version to compile the > >rest of it? I might try downgrading my sources back to the version > >that I last succesfully compiled just to prove it one way or the other > >to myself. > > > > So, been doing some testing. It looks like a -j4 problem with the latest > sources. If I buildworld with -j1 then it compiles with no issues at > all. If I compile r286908 with -j4 then it compiles with no issues at > all. If I try and compile r287155 with -j4 then I get the bus errors. So > I'm not convinced at all that it's hardware related at the moment.Not saying it is. But it still could be a region of CPU cache that never got exercised, or in the right (same) manner. Maybe use a CPU/RAM test program, just to be sure? --Chris> > -- > Matt > _______________________________________________ > freebsd-stable at freebsd.org mailing list > https://lists.freebsd.org/mailman/listinfo/freebsd-stable > To unsubscribe, send any mail to "freebsd-stable-unsubscribe at freebsd.org"