Re: [gentoo-dev] Re: [gentoo-commits] repo/gentoo:master commit in: sys-apps/microcode-ctl/
On Tue, 11 Aug 2015 13:17:10 +0200 hasufell wrote: On 08/11/2015 08:34 AM, Mike Frysinger wrote: commit: 719cc5ef240b766953ddbe1e7a6593f8091eed12 Author: Mike Frysinger vapier AT gentoo DOT org AuthorDate: Tue Aug 11 06:28:16 2015 + Commit: Mike Frysinger vapier AT gentoo DOT org CommitDate: Tue Aug 11 06:34:22 2015 + URL:https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=719cc5ef microcode-ctl: stop installing the init script Updating microcode on the fly is dangerous as it can modify the set of valid instructions. An active example of this is Intel's TSX insns -- the latest microcode push disables the insn on newer CPUs and causes SIGILL when you try to use it. But if you test for the insn before the microcode is updated, it will execute fine. For daemons that launched before the update, they'll find the flag works, and then crash later on when the insn no longer exists. Thus the only safe way to update microcode is at boot time via a builtin initramfs. Details on this operation can be found in #528712#41. I've already asked you twice on the ML why you keep ignoring the standard we set for the commit message summary and pretty much everyone is following except you. Let me remind you that: 1. this is not a standard, but a draft; 2. not all issues are clear right now (e.g. how to reference bugs); 3. it is not approved by the Council; 4. not everyone agrees with these rules anyway. Best regards, Andrew Savchenko pgpX1D8IcKPzO.pgp Description: PGP signature
Re: [gentoo-dev] Re: [gentoo-commits] repo/gentoo:master commit in: sys-apps/microcode-ctl/
On 08/12/2015 12:11 PM, hasufell wrote: On 08/12/2015 08:48 AM, Andrew Savchenko wrote: On Tue, 11 Aug 2015 13:17:10 +0200 hasufell wrote: On 08/11/2015 08:34 AM, Mike Frysinger wrote: commit: 719cc5ef240b766953ddbe1e7a6593f8091eed12 Author: Mike Frysinger vapier AT gentoo DOT org AuthorDate: Tue Aug 11 06:28:16 2015 + Commit: Mike Frysinger vapier AT gentoo DOT org CommitDate: Tue Aug 11 06:34:22 2015 + URL:https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=719cc5ef microcode-ctl: stop installing the init script Updating microcode on the fly is dangerous as it can modify the set of valid instructions. An active example of this is Intel's TSX insns -- the latest microcode push disables the insn on newer CPUs and causes SIGILL when you try to use it. But if you test for the insn before the microcode is updated, it will execute fine. For daemons that launched before the update, they'll find the flag works, and then crash later on when the insn no longer exists. Thus the only safe way to update microcode is at boot time via a builtin initramfs. Details on this operation can be found in #528712#41. I've already asked you twice on the ML why you keep ignoring the standard we set for the commit message summary and pretty much everyone is following except you. Let me remind you that: 1. this is not a standard, but a draft; 2. not all issues are clear right now (e.g. how to reference bugs); 3. it is not approved by the Council; 4. not everyone agrees with these rules anyway. So you want to sabotage consistency and wait until the council approves every minor detail that has been worked out by the community? These things WERE discussed (half a year ago or more) and there was consensus. Things that are still not clear (such as referencing bug reports) were not added to the wiki. Don't mix things up. If we say everyone is allowed to ignore all the rules as long as it repoman-checks, then I don't know why I even keep throwing out discussions, emails and editing the wiki. Commit message format is really not a workflow-limiting thing. I don't know why you complain like that. And vapier has not participated much in that discussion and has not expressed his interest in a different format. So how is anyone supposed to react to that? This is not constructive. And as I see it, even vapier does it now. So everything is good. https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=3581b2c101337dec32c2bd6779db7927ff96732a It was probably related to repoman lacking proper support there. Thanks for your cooperation.
Re: [gentoo-dev] Re: [gentoo-commits] repo/gentoo:master commit in: sys-apps/microcode-ctl/
On 08/12/2015 08:48 AM, Andrew Savchenko wrote: On Tue, 11 Aug 2015 13:17:10 +0200 hasufell wrote: On 08/11/2015 08:34 AM, Mike Frysinger wrote: commit: 719cc5ef240b766953ddbe1e7a6593f8091eed12 Author: Mike Frysinger vapier AT gentoo DOT org AuthorDate: Tue Aug 11 06:28:16 2015 + Commit: Mike Frysinger vapier AT gentoo DOT org CommitDate: Tue Aug 11 06:34:22 2015 + URL:https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=719cc5ef microcode-ctl: stop installing the init script Updating microcode on the fly is dangerous as it can modify the set of valid instructions. An active example of this is Intel's TSX insns -- the latest microcode push disables the insn on newer CPUs and causes SIGILL when you try to use it. But if you test for the insn before the microcode is updated, it will execute fine. For daemons that launched before the update, they'll find the flag works, and then crash later on when the insn no longer exists. Thus the only safe way to update microcode is at boot time via a builtin initramfs. Details on this operation can be found in #528712#41. I've already asked you twice on the ML why you keep ignoring the standard we set for the commit message summary and pretty much everyone is following except you. Let me remind you that: 1. this is not a standard, but a draft; 2. not all issues are clear right now (e.g. how to reference bugs); 3. it is not approved by the Council; 4. not everyone agrees with these rules anyway. So you want to sabotage consistency and wait until the council approves every minor detail that has been worked out by the community? These things WERE discussed (half a year ago or more) and there was consensus. Things that are still not clear (such as referencing bug reports) were not added to the wiki. Don't mix things up. If we say everyone is allowed to ignore all the rules as long as it repoman-checks, then I don't know why I even keep throwing out discussions, emails and editing the wiki. Commit message format is really not a workflow-limiting thing. I don't know why you complain like that. And vapier has not participated much in that discussion and has not expressed his interest in a different format. So how is anyone supposed to react to that? This is not constructive.
[gentoo-dev] Re: [gentoo-commits] repo/gentoo:master commit in: sys-apps/microcode-ctl/
On 08/11/2015 08:34 AM, Mike Frysinger wrote: commit: 719cc5ef240b766953ddbe1e7a6593f8091eed12 Author: Mike Frysinger vapier AT gentoo DOT org AuthorDate: Tue Aug 11 06:28:16 2015 + Commit: Mike Frysinger vapier AT gentoo DOT org CommitDate: Tue Aug 11 06:34:22 2015 + URL:https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=719cc5ef microcode-ctl: stop installing the init script Updating microcode on the fly is dangerous as it can modify the set of valid instructions. An active example of this is Intel's TSX insns -- the latest microcode push disables the insn on newer CPUs and causes SIGILL when you try to use it. But if you test for the insn before the microcode is updated, it will execute fine. For daemons that launched before the update, they'll find the flag works, and then crash later on when the insn no longer exists. Thus the only safe way to update microcode is at boot time via a builtin initramfs. Details on this operation can be found in #528712#41. I've already asked you twice on the ML why you keep ignoring the standard we set for the commit message summary and pretty much everyone is following except you. Do you ignore us on purpose?