The configure option has explicitly thrown an error for more than a year now,
and it happens that the remaining way to still forcefully use it has been
broken for more than 8 months.
Having a template parameter conflict with a global name is terribly
inconvenient, so let's try to avoid that by renaming the 'RefPtr'
template parameter to something else.
Leave a typedef for compatibility. nsVariant will be defined as a
separate class in the next patch.
Also, remove an obsolete comment and fix some whitespace.
|getReportsForThisProcess| differs from |getReports| in that it is limited to current process and is synchronous. When asynchronous memory reporters are added the function will no longer be able tobe synchronous. There isn't much utility in only measuring the current process, so we can remove the function and switch existing users to |getReports|.
The calculation of |explicit| relies on the synchronous |getReportsForThisProcess|, once we have asynchronous reporters this will no longer work. As it is currently referenced in the about::memory tests we can just remove it.
This class can be used instead of raw pointer for a sound leaking-by-default
behavior. Also it could take advantage of move semantic check in the future.
--HG--
extra : source : 6bf72b4eaa92b13d42b547d8aeee677489a4d3e2
Using forget() to extract mMessage from MessageElement ends up going
from nsCOMPtr<T> to already_AddRefed<T> to nsCOMPtr<T>. For the second
step, the compiler can't tell that the already_AddRefed<T> came from a
canonical nsCOMPtr, so it calls Assert_NoQueryNeeded() in debug
builds. This in turn causes a QI, which does an AddRef. That is bad
because we're not on the main thread, and mMessage is
main-thread-only, so we get an assertion.
This patch works around that by using swap directly between two
nsCOMPtr<>, which avoids the Assert_NoQueryNeeded().
I called the method "swapMessage" rather than "swap" to emphasize that
we are not swapping the whole MessageElement, but just one part of
it. I find the existing forget() name to be confusing.
We don't need to reinterpret_cast when casting from void*, and we don't
need to reinterpret_cast when we're casting up and down a class
hierarchy. static_cast takes care of those cases just fine, and doesn't
scare the reader into thinking that nsTHashtable is doing something
unusual.
This class can be used instead of raw pointer for a sound leaking-by-default
behavior. Also it could take advantage of move semantic check in the future.
--HG--
extra : source : 47cd2c22bafc8d4bb1c7e1dce3b45517aaec199f
This changeset replaces all of the
// char16_t[]
optional bytes someProperty = 1;
one- and two-byte string properties in the CoreDump.proto protobuf definition
file with:
oneof {
// char16_t[]
bytes someProperty = 1;
uint64 somePropertyRef = 2;
}
The first time the N^th unique string is serialized, then someProperty is used
and the full string is serialized in the protobuf message. All following times
that string is serialized, somePropertyRef is used and its value is N.
Among the other things, this also changes JS::ubi::Edge::name from a raw pointer
with commented rules about who does or doesn't own and should and shouldn't free
the raw pointer to a UniquePtr that enforces those rules rather than relying on
developers reading and obeying the rules in the comments.
