Is Eve the host machine in this context? Or is this building an image for Eve?

This seems bisectable?

@scollyer: It's an image for Eve.

After a git pull and gclient sync I've been getting the following error instead:

[13096/62564] CXX obj/third_party/swiftshader/src/OpenGL/libEGL/swiftshader_libEGL/Surface.o
FAILED: obj/third_party/swiftshader/src/OpenGL/libEGL/swiftshader_libEGL/Surface.o 
/usr/local/google/home/jordynass/chromium/.cros_cache/common/goma+2/gomacc x86_64-cros-linux-gnu-clang++ -B/usr/local/google/home/jordynass/chromium/.cros_cache/chrome-sdk/tarballs/eve+10503.0.0+target_toolchain/usr/x86_64-pc-linux-gnu/x86_64-cros-linux-gnu/binutils-bin/2.27.0-gold -Wno-unknown-warning-option -MMD -MF obj/third_party/swiftshader/src/OpenGL/libEGL/swiftshader_libEGL/Surface.o.d -DV8_DEPRECATION_WARNINGS -DUSE_UDEV -DUSE_AURA=1 -DUSE_NSS_CERTS=1 -DUSE_OZONE=1 -DFULL_SAFE_BROWSING -DSAFE_BROWSING_CSD -DSAFE_BROWSING_DB_LOCAL -DOFFICIAL_BUILD -DCHROMIUM_BUILD -DFIELDTRIAL_TESTING_ENABLED -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -DCR_CLANG_REVISION=\"328575-1\" -DOS_CHROMEOS -DNDEBUG -DNVALGRIND -DDYNAMIC_ANNOTATIONS_ENABLED=0 -D__STDC_CONSTANT_MACROS -D__STDC_LIMIT_MACROS -DNO_SANITIZE_FUNCTION=__attribute__\(\(no_sanitize\(\"function\"\)\)\) -DANGLE_DISABLE_TRACE -DNDEBUG -DEGL_EGLEXT_PROTOTYPES -DEGLAPI=__attribute__\(\(visibility\(\"default\"\)\)\)\ __attribute__\(\(no_sanitize\(\"function\"\)\)\) -I../../third_party/swiftshader/include -I../../third_party/swiftshader/src -I../../third_party/swiftshader/src/OpenGL -I../.. -Igen -fno-strict-aliasing -funwind-tables -fPIC -pipe -B../../third_party/binutils/Linux_x64/Release/bin -pthread -fcolor-diagnostics -no-canonical-prefixes -m64 -march=x86-64 -O2 -fno-ident -fdata-sections -ffunction-sections -fno-omit-frame-pointer -g2 -gsplit-dwarf -ggnu-pubnames -fvisibility=hidden -Wheader-hygiene -Wstring-conversion -Wtautological-overlap-compare -Wall -Wno-unused-variable -Wno-missing-field-initializers -Wno-unused-parameter -Wno-c++11-narrowing -Wno-covered-switch-default -Wno-unneeded-internal-declaration -Wno-inconsistent-missing-override -Wno-undefined-var-template -Wno-nonportable-include-path -Wno-address-of-packed-member -Wno-unused-lambda-capture -Wno-user-defined-warnings -Wno-enum-compare-switch -Wno-null-pointer-arithmetic -Wno-ignored-pragma-optimize -std=c++11 -fno-exceptions -fno-operator-names -ffunction-sections -fdata-sections -fomit-frame-pointer -Os -m64 -fPIC -march=x86-64 -mtune=generic -Wno-sign-compare -std=gnu++14 -fno-exceptions -fno-rtti --sysroot=../../../.cros_cache/chrome-sdk/tarballs/eve+10503.0.0+sysroot_chromeos-base_chromeos-chrome.tar.xz -fvisibility-inlines-hidden -pipe -pipe -pipe -march=corei7 -fno-split-dwarf-inlining -fdebug-info-for-profiling -D__google_stl_debug_vector=1 -Wno-unknown-warning-option -stdlib=libc++ -Wno-inline-asm -c ../../third_party/swiftshader/src/OpenGL/libEGL/Surface.cpp -o obj/third_party/swiftshader/src/OpenGL/libEGL/swiftshader_libEGL/Surface.o
In file included from ../../third_party/swiftshader/src/OpenGL/libEGL/Surface.cpp:30:
../../third_party/swiftshader/src/Main/libX11.hpp:19:10: fatal error: 'X11/Xlib.h' file not found
#include <X11/Xlib.h>
         ^~~~~~~~~~~~
1 error generated.

Both last night and this morning I has been spitting out 2 errors continuously:

/tmp/ccVxXu9y.s:513: Error: bad register name `%nacl:(%r15,%rbx)'
/tmp/ccVxXu9y.s:524: Error: bad register name `%nacl:(%r15,%r12)'
/tmp/ccVxXu9y.s:527: Error: bad register name `%nacl:(%r15,%rax)'
/tmp/ccVxXu9y.s:548: Error: no such instruction: `naclaspq $40,%r15'
/tmp/ccVxXu9y.s:550: Error: no such instruction: `nacljmp %r11d,%r15'
/tmp/ccVxXu9y.s:589: Error: no such instruction: `naclsspq $40,%r15'
/tmp/ccVxXu9y.s:601: Error: bad register name `%nacl:116(%r15,%rbx)'
/tmp/ccVxXu9y.s:615: Error: bad register name `%nacl:120(%r15,%rbx)'
/tmp/ccVxXu9y.s:621: Error: bad register name `%nacl:116(%r15,%rbx)'
/tmp/ccVxXu9y.s:626: Error: bad register name `%nacl:120(%r15,%rbx)'
/tmp/ccVxXu9y.s:632: Error: no such instruction: `naclaspq $40,%r15'
/tmp/ccVxXu9y.s:637: Error: no such instruction: `nacljmp %r11d,%r15'
/tmp/ccVxXu9y.s:647: Error: bad register name `%nacl:4(%r15,%rbx)'
/tmp/ccVxXu9y.s:660: Error: bad register name `%nacl:4(%r15,%rbx)'
/tmp/ccVxXu9y.s:698: Error: bad register name `%nacl:4(%r15,%rbx)'
/tmp/ccVxXu9y.s:702: Error: bad register name `%nacl:(%r15,%rbx)'
/tmp/ccVxXu9y.s:703: Error: bad register name `%nacl:116(%r15,%rbx)'
/tmp/ccVxXu9y.s:704: Error: bad register name `%nacl:120(%r15,%rbx)'
/tmp/ccVxXu9y.s:708: Error: bad register name `%nacl:4(%r15,%rax)'
/tmp/ccVxXu9y.s:709: Error: bad register name `%nacl:128(%r15,%rbx)'
/tmp/ccVxXu9y.s:710: Error: bad register name `%nacl:129(%r15,%rbx)'
/tmp/ccVxXu9y.s:717: Error: bad register name `%nacl:132(%r15,%rbx)'
/tmp/ccVxXu9y.s:724: Error: bad register name `%nacl:124(%r15,%rbx)'
/tmp/ccVxXu9y.s:732: Error: no such instruction: `nacljmp %r11d,%r15'
/tmp/ccVxXu9y.s:746: Error: no such instruction: `naclsspq $24,%r15'
/tmp/ccVxXu9y.s:764: Error: no such instruction: `naclaspq $24,%r15'
/tmp/ccVxXu9y.s:766: Error: no such instruction: `nacljmp %r11d,%r15'
/tmp/ccVxXu9y.s:798: Error: bad register name `%nacl:4(%r15,%rbx)'
/tmp/ccVxXu9y.s:802: Error: bad register name `%nacl:(%r15,%rbx)'

Correction: I've been building the Chrome binary for Eve (not the full image)

This doesn't seem to be related to infrastructure. 

TL;DR: I've determined the cause, but could use help affecting a fix via GN.

The core issue is that the SimpleChrome (cros chrome-sdk) build is forcing linux_use_bundled_binutils to true [1].  This causes the following flag to be passed to the compiler for all targets:

-B../../third_party/binutils/Linux_x64/Release/bin

For ppapi targets, my assumption is that linux_use_bundled_binutils should be ignored, but my gn-foo is not up to snuff to achieve that.  I added a check for current_cpu != "pnacl" in the compiler config [2], but some of the ppapi targets (tests it looks like) are not pnacl cpu.

[1] https://cs.chromium.org/chromium/src/third_party/chromite/cli/cros/cros_chrome_sdk.py?type=cs&q=linux_use_bundled_binutils&sq=package:chromium&g=0&l=919

[2] https://cs.chromium.org/chromium/src/build/config/compiler/BUILD.gn?q=linux_use_bundled_binutils&sq=package:chromium&g=0&l=56

this was done back here:
  https://chromium-review.googlesource.com/353027

i'm not familiar with what the linux_use_bundled_binutils knob does in general to comment further

For host builds (ie: not for things that are going to run on the target machine), we use the linker and binutils that comes with Chrome. We want to make sure that the system binutils is not used.

Maybe ask for help from dpranke@ for your GN needs?

It's possible that the problem in #c12 is the core issue, but that code hasn't changed in a long time, so I don't know why it would've started failing last night. Can you post a link to the compile error specifically?

It seems plausible that you wouldn't want the bundled binutils for the pnacl toolchains. The problem with setting linux_use_bundled_binutils explicitly in args.gn like we're doing is that it'll override the default value logic in //build/config/compiler/BUILD.gn, so it'll be set to true even for non-linux builds.

There are a few things we could try that might work around this. One would be to modify the nacl toolchains to explicitly pass linux_use_bundled_binutils = false in toolchain_args. Another would be to modify line 454 of //build/config/compiler/BUILD.gn to explicitly check for is_linux or target_cpu, but that latter check seems less likely to be right. A third would be to delete the line setting the arg from cros_sdk, because maybe we don't actually need it now.

However, without actually seeing the full error and understanding why jordynass@ is hitting it now but we're not seeing it elsewhere, I'm reluctant to endorse changing anything.

Thank you for the suggestions, Dirk!  I'll take a look at these and see what is feasible.

For context on the issue, my guess is this particular set of targets never built successfully in the 'cros chrome-sdk' shell.  Most folks just build the 'chrome' target.  I don't believe any buildbots build all targets in this environment, so I've attached the log output of a trunk build with no modifications.

(sdk eve R70-10916.0.0) jhawkins@jhawkins003 /usr/local/google/src/chromium/src $ autoninja -C out_eve/Release >& eve_build.log

Deleted: eve_build.log 3.3 MB

eve_build.log 3.3 MB View Download

Note: Per the Simple Chrome docs, we haven't made any effort to support targets other than chrome, chrome_sandbox, nacl_helper and chromeos_preflight:

https://chromium.googlesource.com/chromiumos/docs/+/master/simple_chrome_workflow.md#build-chrome

If we need/want to support ppapi or other targets, and we are able to do so, please add a note to the documentation, perhaps with a description of the use case.

(Note: I am not advocating for or against such support, just adding historical context and requesting documentation for the future in case we regress).

Thanks!

Steven: Understood, and will do.  I've had several nooglers get hung up by trying to build all targets (as opposed to just Chrome), so I'm hopeful we can remove that pitfall (and ultimately have the build structured correctly for this flow).