Re: Exact rooting is now enabled on desktop
On 01/20/2014 12:53 PM, Cameron Kaiser wrote: On 1/17/14 1:24 PM, Terrence Cole wrote: SpiderMonkey will no longer scan the stack for live roots; you must now wrap all GC thing pointers that live on the stack across a GC in the JS::Rooted template for SpiderMonkey to see them. See the comments in js/public/RootingAPI.h and the guide on the developer wiki[1] for further details. All currently existing instances in the browser build have been updated; the hazard analysis build on TBPL, SM(Hf), will catch any future pushes that break this invariant. Great work! My only question: do JIT backends need to be adjusted for this? I assume not, but I want to make sure. You are correct, there should be no extra work. The codegen bits should be running in a context that cannot GC. Cameron Kaiser ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
On 1/17/14 1:24 PM, Terrence Cole wrote: SpiderMonkey will no longer scan the stack for live roots; you must now wrap all GC thing pointers that live on the stack across a GC in the JS::Rooted template for SpiderMonkey to see them. See the comments in js/public/RootingAPI.h and the guide on the developer wiki[1] for further details. All currently existing instances in the browser build have been updated; the hazard analysis build on TBPL, SM(Hf), will catch any future pushes that break this invariant. Great work! My only question: do JIT backends need to be adjusted for this? I assume not, but I want to make sure. Cameron Kaiser ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
On Fri, Jan 17, 2014 at 3:22 PM, Jim Blandy j...@mozilla.com wrote: I've never heard of a major project escaping from conservative GC once it had entered that state of sin; nor have I heard of anyone implementing a moving collector after starting with a non-moving collector. There's no better way to answer the question has X been done? than proclaiming loudly on the internet that X has never been done! I did this by quoting Jim in a blog post [1] and was promptly informed that both Racket and Mono have done this. But who cares? This is a fantastic achievement, and I eagerly anticipate the changes it will enable. Nick [1] https://blog.mozilla.org/nnethercote/2014/01/20/a-big-step-towards-generational-and-compacting-gc/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
On 1/18/14, 9:04 AM, Terrence Cole wrote: Great question! We have a tool called GC zeal in builds with --enable-gc-zeal and in all debug builds unconditionally. It adds a small runtime overhead, but gives us fine-grained control over when GC's happen and adds several verification modes for debugging specific problems. I was under the impression that nightlys had this enabled, but I am not seeing it there now. Are there (inexpensive) runtime sanity checks that could be enabled in all Nightly release builds? The Nightly channel already has extra checks enabled that make it inappropriate for benchmarking, like --enable-profiling and some ref counting checks that will MOZ_CRASH. chris ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
On 1/17/2014 4:24 PM, Terrence Cole wrote: Exact stack rooting is now enabled by default on desktop builds of firefox. Does this mean that the moving GC is also enabled, or is that a later step? If we see an increase in the crash rate for nightly builds, is it likely that they will share a signature, or will this mainly expose itself as memory corruption, which typically don't have useful crash stacks/signatures? Are there ways for users who are seeing crashes to turn on extra debugging (memory poisoning in the GC, for example) in order to make crashes more reliable? The stability team will be watching crash-stats closely over the next week. --BDS ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
On 01/18/2014 07:08 AM, Benjamin Smedberg wrote: On 1/17/2014 4:24 PM, Terrence Cole wrote: Exact stack rooting is now enabled by default on desktop builds of firefox. Does this mean that the moving GC is also enabled, or is that a later step? No, moving GC is a later step. We are targeting moving GC to land in the next cycle so that we can more easily separate the problems. If we see an increase in the crash rate for nightly builds, is it likely that they will share a signature, or will this mainly expose itself as memory corruption, which typically don't have useful crash stacks/signatures? The second, although stacks will probably have a relatively tight cluster too. Specifically, a crash here will manifest as use-after-free, generally through some specific, local GC pointer variable. Are there ways for users who are seeing crashes to turn on extra debugging (memory poisoning in the GC, for example) in order to make crashes more reliable? Great question! We have a tool called GC zeal in builds with --enable-gc-zeal and in all debug builds unconditionally. It adds a small runtime overhead, but gives us fine-grained control over when GC's happen and adds several verification modes for debugging specific problems. I was under the impression that nightlys had this enabled, but I am not seeing it there now. The stability team will be watching crash-stats closely over the next week. Thanks for keeping an eye out! -Terrence --BDS ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
- Original Message - Great question! We have a tool called GC zeal in builds with --enable-gc-zeal and in all debug builds unconditionally. It adds a small runtime overhead, but gives us fine-grained control over when GC's happen and adds several verification modes for debugging specific problems. I was under the impression that nightlys had this enabled, but I am not seeing it there now. Nightly has compartment checking enabled, which may be what you are thinking of. Andrew ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
Congratulations! This has been an epic journey :-). Rob -- Jtehsauts tshaei dS,o n Wohfy Mdaon yhoaus eanuttehrotraiitny eovni le atrhtohu gthot sf oirng iyvoeu rs ihnesa.rt sS?o Whhei csha iids teoa stiheer :p atroa lsyazye,d 'mYaonu,r sGients uapr,e tfaokreg iyvoeunr, 'm aotr atnod sgaoy ,h o'mGee.t uTph eann dt hwea lmka'n? gBoutt uIp waanndt wyeonut thoo mken.o w ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
On 1/17/14 3:24 PM, Terrence Cole wrote: Exact stack rooting is now enabled by default on desktop builds of firefox. *standing ovation forever* -j ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Exact rooting is now enabled on desktop
On 01/17/2014 01:24 PM, Terrence Cole wrote: Exact stack rooting is now enabled by default on desktop builds of firefox. I've never heard of a major project escaping from conservative GC once it had entered that state of sin; nor have I heard of anyone implementing a moving collector after starting with a non-moving collector. So, doing *both* is impressive. I hope it pays off big! ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform