This is portable back to older Android NDK APIs.
Seems to work fine.
Change-Id: I1f121f372d376909011ffd7063b73cbe50402a5f
Reviewed-on: https://skia-review.googlesource.com/40688
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This adds a fallback backtracer for use on Android where <execinfo.h>
ins't present, instead using <unwind.h> to unwind and <dlfcn.h> to
lookup function names and addresses.
lockf() wasn't available until NDK API 24, so I've just no-op'd file
locking on older targets. I tried switching from lockf() to flock(),
but flock() didn't see to _do_ anything, neither on Android nor on my
Mac laptop. I think I should be able to use the lower-level fcntl()
APIs to restore file locking uniformly in a follow-up. The upshot is
until then, we'll have interlaced logs and stack traces on Android
devices unless you set ndk_api=24 in GN.
We need to add a couple build flags to make backtraces useful:
* -funwind-tables makes the call to _Unwind_Backtrace() actually
traverse the call stack. This is a small extra binary size cost.
* -rdynamic makes symbols linked into the main executable visible
to dladdr(). We do this on Linux already for the same reason.
Here's an example where I made aaxfermodes call SK_ABORT():
650 ok, 1 crashed
caught signal SIGABRT while running 'aaxfermodes'
0x76ed936288 [unknown]+308
0x76eec014e0 [unknown]+510811706592
0x76ed367b2c tgkill+8
0x76ed364f50 pthread_kill+68
0x76ed31ff5c raise+28
0x76ed318814 abort+56
0x76edebd070 sk_out_of_memory()+12
0x76ed99f664 AAXfermodesGM::draw_pass(SkCanvas*, AAXfermodesGM::DrawingPass)+96
0x76ed99f4e4 AAXfermodesGM::onDraw(SkCanvas*)+36
0x76ed9e8550 skiagm::GM::drawContent(SkCanvas*)+224
0x76ed9e82ac skiagm::GM::draw(SkCanvas*)+288
0x76ed93b10c GMStream::GMSrc::draw(SkCanvas*)+96
0x76ed937b08 SWDst::draw(Src*)+284
0x76ed936ca0 [unknown]+112
0x76ed939b4c ForkEngine::spawn(std::function<Status ()>)+88
0x76ed934d00 main+2200
0x76ed316598 __libc_init+92
0x76ed93434c [unknown]+510791992140
Change-Id: Ica4849d99a3b97f48d778f4c15a7fa36275b8133
Reviewed-on: https://skia-review.googlesource.com/40802
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Change-Id: Ie69421ed36d596cb8c40f9858fac370ceba18c87
Reviewed-on: https://skia-review.googlesource.com/40022
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This makes Engines (task execution strategies: serial, thread, fork)
pluggable just like most of the rest of ok. It removes the thread and
process limits, as I find myself rarely caring about what they are
exactly. Instead of limiting to num-cores, we just allow any number of
concurrent threads, and any number of concurrent child processes subject
to OS limitations.
Change-Id: Icef49d86818fe9a4b7380efb60e73e40bc2e6b73
Reviewed-on: https://skia-review.googlesource.com/27140
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit 9ac801c12c.
Reason for revert: breaks threaded mode.
Original change's description:
> ok, exit() child processes instead of _exit()
>
> We want the atexit() hook to print deferred logs (crash dumps, etc.) only when the main, parent process exits. Ending the child processes with _exit() instead of exit() avoids calling atexit() hooks, but also global destructors. This is more complex than we need: we can just print before main() exits. Only the main, parent process gets there.
>
> This now runs each child process' global destructors, which makes trace.json "work": we get a trace of whichever child process finishes last. It's a start.
>
> Change-Id: I0cc2b12592f0f79e73a43a160b9fd06dba1fee25
> Reviewed-on: https://skia-review.googlesource.com/26800
> Commit-Queue: Mike Klein <mtklein@chromium.org>
> Reviewed-by: Brian Osman <brianosman@google.com>
TBR=mtklein@chromium.org,brianosman@google.com
Change-Id: I695dddaab3b5a51e4698bb7222fc723b544a1d64
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/26943
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
We want the atexit() hook to print deferred logs (crash dumps, etc.) only when the main, parent process exits. Ending the child processes with _exit() instead of exit() avoids calling atexit() hooks, but also global destructors. This is more complex than we need: we can just print before main() exits. Only the main, parent process gets there.
This now runs each child process' global destructors, which makes trace.json "work": we get a trace of whichever child process finishes last. It's a start.
Change-Id: I0cc2b12592f0f79e73a43a160b9fd06dba1fee25
Reviewed-on: https://skia-review.googlesource.com/26800
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Brian Osman <brianosman@google.com>
SkLights.h pulls in a bunch of other headers and is not needed (fwdecl
works fine).
Change-Id: I3ed97cd7861e51dcb7cfa7950a97b420dbc6fbfb
TBR=reed@google.com
Reviewed-on: https://skia-review.googlesource.com/15143
Commit-Queue: Florin Malita <fmalita@chromium.org>
Reviewed-by: Florin Malita <fmalita@chromium.org>
This reverts commit 9ff301bf91.
Reason for revert: need to update G3, Flutter.
Original change's description:
> Remove SkLights include from SkCanvas.h
>
> SkLights.h pulls in a bunch of other headers and is not needed (fwdecl
> works fine).
>
> Change-Id: Id2d7176eb3bf4609f72f46d513eebf59318f542f
> Reviewed-on: https://skia-review.googlesource.com/14904
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Florin Malita <fmalita@chromium.org>
>
TBR=mtklein@google.com,fmalita@chromium.org,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I4799ad5b31aaeaf529c8b912bbe09aa8869a5e6c
Reviewed-on: https://skia-review.googlesource.com/15107
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Florin Malita <fmalita@chromium.org>
SkLights.h pulls in a bunch of other headers and is not needed (fwdecl
works fine).
Change-Id: Id2d7176eb3bf4609f72f46d513eebf59318f542f
Reviewed-on: https://skia-review.googlesource.com/14904
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Florin Malita <fmalita@chromium.org>
wait_for(delta) is wait_until(steady_clock::now() + delta) under the
hood, so using wait_for() like this implies an extra call to now() that
we can avoid by using wait_until().
We can hoist that call out and just do it once... the past stays the past.
This is not super important. Just noticed while profiling. It's nice
to keep the overhead of the ok tool down so the real work can show. :)
Change-Id: I89d25a800b63ebcfc229b5b3aa3f2dd621f4e7b4
Reviewed-on: https://skia-review.googlesource.com/14480
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
-m, -s, and -w are now vias.
No real need to advertise -h given that if you just run ok with no
arguments it prints the help, but to be friendly accept -h and --help.
Change-Id: Id23936106cfea7d670cf0eb9773a5851055576f1
Reviewed-on: https://skia-review.googlesource.com/13254
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This refactors most features out of ok's core into vias:
-w --> a .png dumping via, "png", opening the door to other types
-m/-s --> a filtering via "filter"
Everything now can print a brief help message too.
Change-Id: I9e653aab98fd57182a6d458c7a80052130980284
Reviewed-on: https://skia-review.googlesource.com/10509
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Just as deferred and locked crash logging makes crashes easier to read,
so does deferred and locked failure logging make failures easier to read.
Change-Id: I71578d61b0056f8d7e692149762def1f155c0387
Reviewed-on: https://skia-review.googlesource.com/10280
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Plenty TODO remaining.
This can actually kind of run without a test reporter or GrContext:
$ out/ok test
784 ok, 56 crashed
... lots of stack traces ...
Most tests don't use the reporter unless they're going to fail.
Change-Id: I7333e2c63ade5e671ebf60022d19390f1fc1c93a
Reviewed-on: https://skia-review.googlesource.com/10201
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This is mostly a demo, and to make sure it's easy.
If I'm thinking right, other non-ct options should Just Work.
Change-Id: I295db0fa04921ccdd766e1870e367594ca802462
Reviewed-on: https://skia-review.googlesource.com/10190
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
lockf() is a simple way to prevent interlaced stack traces when there
are concurrent crashes. After a crashing process rethrows its signal
and dies for real, the OS unlocks the file for other processes.
I tested this by making SkCanvas::drawRRect() crash on Linux:
20-odd GM crashes with interlaced stack traces before, none after.
Change-Id: I99930756b8c85c552eef7c3a77778e4c00d34c42
Reviewed-on: https://skia-review.googlesource.com/10177
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This makes everything a lot more like DM, for the same reason:
it's the best way to make Vias work.
Instead of exposing a canvas, Dsts take a Src to draw. Vias still are
Dsts that wrap Dsts. They do their internal work in draw() then pass a
proxy Src encapsulating that work to the next Dst's draw().
A little refactoring in ok.cpp allows arbitrary chains of Vias.
I removed the guarantee that Src methods are called in strict order.
It's easy enough to make each Src initialize itself as needed.
I moved the .png encoding back to ok.cpp. It seemed weird for Dsts to
have to think about files and paths. One day Dst will want a data()
method for non-image output (.pdf, .skp), and then we'll want ok.cpp to
be the one to coordinate what to write where.
Change-Id: Id4a3674b2d05aef2b5f10e0077df0a8407c07b61
Reviewed-on: https://skia-review.googlesource.com/10175
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Not sure if these simple Src/Dst interfaces will last.
Vias are a little tricky, and some may be impossible.
Change-Id: I42d19b1ee74b51a830bb781f25a888c0b32ba98c
Reviewed-on: https://skia-review.googlesource.com/10174
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Now ok.cpp handles only the high level coordination of Srcs and Dsts,
without having to know or care what they are.
Some minor refactoring to things like Options.
Change-Id: I02df890b26d6d069e980a125b6a1ce1a7067b900
Reviewed-on: https://skia-review.googlesource.com/10173
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Change-Id: I47ac01f0c2c0f2f7b925de09c18d3c8265398c8a
Reviewed-on: https://skia-review.googlesource.com/10117
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
The 'dont_clip_to_layer' GM does not balance its layers. It calls
saveLayer() 3x but restore() only 2x. This may be a bug in the GM
itself, but I'm not sure so I haven't changed it here.
If ok is writing .pngs, the surface passes ownership of its buffer to an
image snapshot, simply marking it as immutable. Then, when the surface
is destroyed later, it destroys its inner canvas, which restores its
save stack to zero, actually doing some drawing in the case of
unbalanced saveLayer()s. We then call notifyPixelsChanged() and hit an
assert saying "you just wrote some pixels but this buffer was marked
immutable."
DM doesn't show this problem because it's doesn't really use surfaces
and images, just bitmaps. There's no ownership handoff and nothing is
ever immutable, so the condition triggering the assert never comes up.
I'm not really sure where we want to say is the bug:
- SkCanvas can draw in its destructor?
- SkSurface doesn't restore to zero before snapping an image?
- that dont_clip_to_layer should call restore three times?
In any case, this guards against it in ok.
I was using this as a convenient crash to help figure out how to best
save and print stack traces, but now that I've got that worked out we
might as well fix this.
Change-Id: Id6d397f534dd1b50219e0d3078c989a4910883a6
Reviewed-on: https://skia-review.googlesource.com/10140
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This seems to work pretty nicely for each engine in the expected case of
few crashes. The serial and thread engines just dump the first crash
stack to stderr before dying, while the fork engine saves all crashes to
a temporary file, then prints that to stderr once everything's finished.
I'm not sold on this TLS solution as being the best way to know what
was running when we crashed, but it's better than printing nothing.
Change-Id: I0aca66529301b1ad9bd51ec728848817586c606d
Reviewed-on: https://skia-review.googlesource.com/10102
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This runs much faster. Very good idea.
Change-Id: I088aa9588c069a17e4745be55c2397114ee8a2bc
Reviewed-on: https://skia-review.googlesource.com/10053
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Change-Id: I9d01656a9d9b7aa3ab352dd4c168b26da620a903
Reviewed-on: https://skia-review.googlesource.com/9978
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
