Closed Bug 1102541 Opened 11 years ago Closed 11 years ago

JS::ubi::RootList should have a method for manually adding ubi::Nodes as roots

Categories

(Core :: JavaScript Engine, defect)

x86
macOS
defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla37

People

(Reporter: fitzgen, Assigned: fitzgen)

References

Details

Attachments

(1 file, 2 obsolete files)

If some ubi::Node isn't technically a GC root (like a debuggee global), we might want to pretend it is anyways, and it would be nice if RootList had an API to support that.
Depends on: 1108149
Attached patch add-root-to-root-list.patch (obsolete) — Splinter Review
Attachment #8534542 - Flags: review?(shu)
Comment on attachment 8534542 [details] [diff] [review] add-root-to-root-list.patch Review of attachment 8534542 [details] [diff] [review]: ----------------------------------------------------------------- More straightforward, nice. ::: js/src/vm/Debugger.cpp @@ +3575,5 @@ > + /* > + * Ensure that each of our debuggee globals are in the root list. > + */ > + for (GlobalObjectSet::Range r = dbg->allDebuggees(); !r.empty(); r.popFront()) { > + if (!rootList.addRoot(JS::ubi::Node(static_cast<JSObject *>(r.front())))) Do you need the static_cast? @@ +3578,5 @@ > + for (GlobalObjectSet::Range r = dbg->allDebuggees(); !r.empty(); r.popFront()) { > + if (!rootList.addRoot(JS::ubi::Node(static_cast<JSObject *>(r.front())))) > + return false; > + } > + Hm, why not put this logic into RootList::init(HandleObject)? That already expects a Debugger object and operates on its debuggee zones, so might as well just add the globals. I don't feel that strongly though. ::: js/src/vm/DebuggerMemory.cpp @@ +779,5 @@ > return false; > > + // Add all debuggee globals as roots. > + for (GlobalObjectSet::Range r = dbg->allDebuggees(); !r.empty(); r.popFront()) { > + if (!rootList.addRoot(JS::ubi::Node(static_cast<JSObject *>(r.front())))) Ditto on if you need this static_cast.
Attachment #8534542 - Flags: review?(shu) → review+
(In reply to Shu-yu Guo [:shu] from comment #2) > ::: js/src/vm/Debugger.cpp > @@ +3575,5 @@ > > + /* > > + * Ensure that each of our debuggee globals are in the root list. > > + */ > > + for (GlobalObjectSet::Range r = dbg->allDebuggees(); !r.empty(); r.popFront()) { > > + if (!rootList.addRoot(JS::ubi::Node(static_cast<JSObject *>(r.front())))) > > Do you need the static_cast? Yup, I get compiler errors without it here and in the other location. > @@ +3578,5 @@ > > + for (GlobalObjectSet::Range r = dbg->allDebuggees(); !r.empty(); r.popFront()) { > > + if (!rootList.addRoot(JS::ubi::Node(static_cast<JSObject *>(r.front())))) > > + return false; > > + } > > + > > Hm, why not put this logic into RootList::init(HandleObject)? That already > expects a Debugger object and operates on its debuggee zones, so might as > well just add the globals. I don't feel that strongly though. Good idea, will do.
Attached patch add-root-to-root-list.patch (obsolete) — Splinter Review
Updated based on review comments. Carrying over shu's r+. Tree is closed right now, so can't land it yet.
Attachment #8534542 - Attachment is obsolete: true
Attachment #8535222 - Flags: review+
I don't understand why the compiler on linux doesn't like this... It is complaining about GlobalObjectSet: > 12:39:47 INFO - /builds/slave/m-in-lx-d-00000000000000000000/build/src/js/src/vm/UbiNode.cpp: In member function 'bool JS::ubi::RootList::init(JS::HandleObject)': > 12:39:47 INFO - /builds/slave/m-in-lx-d-00000000000000000000/build/src/js/src/vm/UbiNode.cpp:330:10: error: 'GlobalObjectSet' has not been declared But in GlobalObject.h GlobalObjectSet gets defined (http://dxr.mozilla.org/mozilla-central/source/js/src/vm/GlobalObject.h?from=GlobalObjectSet#809): > typedef HashSet<GlobalObject *, DefaultHasher<GlobalObject *>, SystemAllocPolicy> GlobalObjectSet; And in UbiNode.cpp we include that header (http://dxr.mozilla.org/mozilla-central/source/js/src/vm/UbiNode.cpp?from=UbiNode.cpp#24): > #include "vm/GlobalObject.h" Any idea what's up here, shu?
Flags: needinfo?(nfitzgerald) → needinfo?(shu)
GlobalObjectSet is in the js namespace, so maybe js::GlobalObjectSet?
(In reply to Jan de Mooij [:jandem] from comment #8) > GlobalObjectSet is in the js namespace, so maybe js::GlobalObjectSet? Sigh, thanks >_<
Flags: needinfo?(shu)
Updated patch from the green try push.
Attachment #8535222 - Attachment is obsolete: true
Attachment #8537310 - Flags: review+
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla37
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: