The current rust panic hook keeps a string for the crash reporter, and
goes on calling the default rust panic hook, which prints out a crash
stack... when RUST_BOOTSTRAP is set *and* when that works. Notably, on
both mac and Windows, it only really works for local builds, but fails
for debug builds from automation, although on automation itself, we also
do stackwalk from crash minidumps, which alleviates the problem.
Artifact debug builds are affected, though.
More importantly, C++ calls to e.g. MOZ_CRASH have a similar but
different behavior, in that they dump a stack trace on debug builds, by
default (with exceptions, see below for one). The format of those stack
traces is understood by the various fix*stack*py scripts under
tools/rb/, that are used by the various test harnesses both on
automation and locally.
Additionally, the current rust panic hook, as it calls the default rust
panic hook, ends up calling abort() on non-Windows platforms, which ends
up being verbosely redirected to mozalloc_abort per
https://dxr.mozilla.org/mozilla-central/rev/237e4c0633fda8e227b2ab3ab57e417c980a2811/memory/mozalloc/mozalloc_abort.cpp#79
which then calls MOZ_CRASH. Theoretically, /that/ would also print a
stack trace, but doesn't because currently the stack trace printing code
lives in libxul, and MOZ_CRASH only calls it when compiled from
libxul-code, which mozalloc_abort is not part of.
With this change, we make the rust panic handler call back into
MOZ_CRASH directly. This has multiple advantages:
- This is more consistent cross-platforms (Windows is not special
anymore).
- This is more consistent between C++ and rust (stack traces all look
the same, and can all be post-processed by fix*stack*py if need be)
- This is more consistent in behavior, where debug builds will show
those stack traces without caring about environment variables.
- It demangles C++ symbols in rust-initiated stack traces (for some
reason that didn't happen with the rust panic handler)
A few downsides:
- the loss of demangling for some rust symbols.
- the loss of addresses in the stacks, although they're not entirely
useful
- extra empty lines.
The first should be fixable later one. The latter two are arguably
something that should be consistent across C++ and rust, and should be
changed if necessary, independently of this patch.
Depends on D11719
Depends on D11719
Differential Revision: https://phabricator.services.mozilla.com/D11720
--HG--
extra : moz-landing-system : lando
The current rust panic hook keeps a string for the crash reporter, and
goes on calling the default rust panic hook, which prints out a crash
stack... when RUST_BOOTSTRAP is set *and* when that works. Notably, on
both mac and Windows, it only really works for local builds, but fails
for debug builds from automation, although on automation itself, we also
do stackwalk from crash minidumps, which alleviates the problem.
Artifact debug builds are affected, though.
More importantly, C++ calls to e.g. MOZ_CRASH have a similar but
different behavior, in that they dump a stack trace on debug builds, by
default (with exceptions, see below for one). The format of those stack
traces is understood by the various fix*stack*py scripts under
tools/rb/, that are used by the various test harnesses both on
automation and locally.
Additionally, the current rust panic hook, as it calls the default rust
panic hook, ends up calling abort() on non-Windows platforms, which ends
up being verbosely redirected to mozalloc_abort per
https://dxr.mozilla.org/mozilla-central/rev/237e4c0633fda8e227b2ab3ab57e417c980a2811/memory/mozalloc/mozalloc_abort.cpp#79
which then calls MOZ_CRASH. Theoretically, /that/ would also print a
stack trace, but doesn't because currently the stack trace printing code
lives in libxul, and MOZ_CRASH only calls it when compiled from
libxul-code, which mozalloc_abort is not part of.
With this change, we make the rust panic handler call back into
MOZ_CRASH directly. This has multiple advantages:
- This is more consistent cross-platforms (Windows is not special
anymore).
- This is more consistent between C++ and rust (stack traces all look
the same, and can all be post-processed by fix*stack*py if need be)
- This is more consistent in behavior, where debug builds will show
those stack traces without caring about environment variables.
- It demangles C++ symbols in rust-initiated stack traces (for some
reason that didn't happen with the rust panic handler)
A few downsides:
- the loss of demangling for some rust symbols.
- the loss of addresses in the stacks, although they're not entirely
useful
- extra empty lines.
The first should be fixable later one. The latter two are arguably
something that should be consistent across C++ and rust, and should be
changed if necessary, independently of this patch.
Depends on D11719
Differential Revision: https://phabricator.services.mozilla.com/D11720
--HG--
extra : moz-landing-system : lando
The current rust panic hook keeps a string for the crash reporter, and
goes on calling the default rust panic hook, which prints out a crash
stack... when RUST_BOOTSTRAP is set *and* when that works. Notably, on
both mac and Windows, it only really works for local builds, but fails
for debug builds from automation, although on automation itself, we also
do stackwalk from crash minidumps, which alleviates the problem.
Artifact debug builds are affected, though.
More importantly, C++ calls to e.g. MOZ_CRASH have a similar but
different behavior, in that they dump a stack trace on debug builds, by
default (with exceptions, see below for one). The format of those stack
traces is understood by the various fix*stack*py scripts under
tools/rb/, that are used by the various test harnesses both on
automation and locally.
Additionally, the current rust panic hook, as it calls the default rust
panic hook, ends up calling abort() on non-Windows platforms, which ends
up being verbosely redirected to mozalloc_abort per
https://dxr.mozilla.org/mozilla-central/rev/237e4c0633fda8e227b2ab3ab57e417c980a2811/memory/mozalloc/mozalloc_abort.cpp#79
which then calls MOZ_CRASH. Theoretically, /that/ would also print a
stack trace, but doesn't because currently the stack trace printing code
lives in libxul, and MOZ_CRASH only calls it when compiled from
libxul-code, which mozalloc_abort is not part of.
With this change, we make the rust panic handler call back into
MOZ_CRASH directly. This has multiple advantages:
- This is more consistent cross-platforms (Windows is not special
anymore).
- This is more consistent between C++ and rust (stack traces all look
the same, and can all be post-processed by fix*stack*py if need be)
- This is more consistent in behavior, where debug builds will show
those stack traces without caring about environment variables.
- It demangles C++ symbols in rust-initiated stack traces (for some
reason that didn't happen with the rust panic handler)
A few downsides:
- the loss of demangling for some rust symbols.
- the loss of addresses in the stacks, although they're not entirely
useful
- extra empty lines.
The first should be fixable later one. The latter two are arguably
something that should be consistent across C++ and rust, and should be
changed if necessary, independently of this patch.
Depends on D11719
Differential Revision: https://phabricator.services.mozilla.com/D11720
--HG--
extra : moz-landing-system : lando
This is the equivalent of the rustc-workspace-hack used by the rust build to
ensure cargo and RLS see the same set of features for dependencies so that
these dependencies may be reused by invocations of cargo for these two
projects. The trivial crate added specifies the union of the set of
features activated for a particular crate for each time it appears in the
dependency tree so that cargo will understand these dependencies to be
re-usable across cargo implementations. This eliminates re-building jsrust
and some of its dependencies twice, and reduces the number of crates compiled
in the tree by about 90 in testing on linux.
Differential Revision: https://phabricator.services.mozilla.com/D9041
Updating rkv to 0.5 enables us to un-vendor new-ordered-float, as rkv 0.4 is the last crate in the tree that depends on it.
It also enables us to un-vendor version 0.5 of uuid. We previously needed that version because multiple third-party crates depended on it, and we have limited control over third-party sub-dependencies. But rkv 0.4 was the last third-party crate that still depended on version 0.5 of uuid; rkv 0.5 depends on version 0.6 of uuid.
There would still be two internal crates that depend on version 0.5 of uuid: geckodriver and webrender_bindings. But we have more control over internal sub-dependencies, and we can update those two internal crates to depend on version 0.6 of uuid. This patch does so.
To summarize, this patch makes the following changes:
* rkv: 0.4 -> 0.5
* new-ordered-float: un-vendored
* geckodriver: uuid dependency 0.5 -> 0.6
* webrender_bindings: uuid dependency 0.5 -> 0.6
* uuid 0.5: un-vendored
* uuid 0.6: remains in tree
Differential Revision: https://phabricator.services.mozilla.com/D9160
--HG--
extra : moz-landing-system : lando
This is the equivalent of the rustc-workspace-hack used by the rust build to
ensure cargo and RLS see the same set of features for dependencies so that
these dependencies may be reused by invocations of cargo for these two
projects. The trivial crate added specifies the union of the set of
features activated for a particular crate for each time it appears in the
dependency tree so that cargo will understand these dependencies to be
re-usable across cargo implementations. This eliminates re-building jsrust
and some of its dependencies twice, and reduces the number of crates compiled
in the tree by about 90 in testing on linux.
Differential Revision: https://phabricator.services.mozilla.com/D9041
--HG--
extra : moz-landing-system : lando
Implemnt notification backend by Windows Toast API that is from Windows 8+.
Original patch is me and add some features by eoger.
Differential Revision: https://phabricator.services.mozilla.com/D3003
--HG--
extra : rebase_source : 0368f269e9adb2347621500b7c9d62c172a71e39
Implemnt notification backend by Windows Toast API that is from Windows 8+.
Original patch is me and add some features by eoger.
Differential Revision: https://phabricator.services.mozilla.com/D3003
--HG--
extra : rebase_source : 5b73af9480569105c24baa5013e25879cbc68b02
r?njn for the profiler parts
r?jrmuizel for the ELF parsing parts
Depends on D7020
Differential Revision: https://phabricator.services.mozilla.com/D7021
--HG--
extra : moz-landing-system : lando
r?njn for the profiler parts
r?jrmuizel for the ELF parsing parts
Depends on D7020
Differential Revision: https://phabricator.services.mozilla.com/D7021
--HG--
extra : moz-landing-system : lando
This introduces two new crates:
- jsrust, for standalone builds. This crate is compiled into a static library
libjsrust.a, which gets linked into the shared Spidermonkey library when it's
built, or into the static Spidermonkey library otherwise. This is just a
static library wrapping jsrust_shared below.
- jsrust_shared, for Gecko embedding. It just references other Rust
crates actively used in Spidermonkey. It is used to be embedded as part of
a new Rust dependency in Gecko (in gkrust).
--HG--
rename : js/src/wasm/cranelift/Cargo.toml => js/src/rust/Cargo.toml
extra : rebase_source : 84e440e3f669b73776653182cb7b006cc7febb10
extra : histedit_source : 3a67575ff6871b7dc3558c10a0251b73cedb090c
This implements an API in `nsIOSKeyStore.idl` and `OSKeyStore.cpp` to encrypt and decrypt bytes with a key that is stored in the OS key store.
There are two OS adapters in this patch.
Libsecret is used on Linux if available.
The NSS key store is used as fallback if no OS specific key store is implemented.
Differential Revision: https://phabricator.services.mozilla.com/D1858
--HG--
extra : rebase_source : 99d7d646968a46a13ffa61885bb246f6d3e443e4
Now that we've generated an order file of the first N functions invoked
during startup, let's tell the linker about said functions so it can
cluster them appropriately.
When I originally implemented bug 1458161, this is how it was done, but
it was suggested to use a configure-time check. This turned out to not
be great, because the rust compiler changes regularly, and we don't run
the configure tests when the version changes. When people upgraded their
rust compiler to 1.27, the code subsequently failed to build because the
features were still set for the previous version they had installed.
--HG--
extra : rebase_source : 1b5f7a02ad8495d68cd29289f7beea59b8912183
Overall, this makes the whole setup less fragile, and make it work with
LTO in more situations.
--HG--
extra : rebase_source : de968c61dc4ef337fdc28745c202334ac41763cd
We perform, on the binaries we build, a series of check, that are
implemented as half-baked make commands, invoked after linking them.
- check libstdc++ symbol versions to ensure binary compatibility with
a baseline.
- check glibc symbol versions to ensure binary compatibility with a
baseline.
- check that target binaries don't contain text relocations.
- check that libmozglue is linked before libc on android.
- on libxul, check that NSModules are laid out correctly.
- on libxul, check that there is more than one PT_LOAD segment.
Those checks happen to work where they matter, but their setup is
unreliable. For example, the checks for symbol versions are supposed to
work for libclang-plugin on cross osx builds, but in fact, don't,
because the readelf path doesn't exist, and the command doesn't fail in
that case.
So move them all to a standalone script, performing the checks more
thoroughly (especially the NSModules one, where we now also check that
they are all adjacent), and more verbosely.
--HG--
extra : rebase_source : 7072e622e95f363d4a6c3a8e272d3445d998b592
Sylvestre Ledru
Mike Hommey
Dorel Luca
Coroiu Cristina
Nathan Froyd
Andrew Osmond
Chris Manchester
Margareta Eliza Balazs
Myk Melez
Edouard Oger
Ciure Andrei
Makoto Kato
Markus Stange
shindli
Benjamin Bouvier
Tom Ritter
sotaro
Narcis Beleuzu
Daniel Varga
Andreea Pavel
Franziskus Kiefer
Nicholas Nethercote
Stephen A Pohl
Tiberius Oros
Nathan Froyd
Jacek Caban