fae9084ac5 build: Skip i686 build by default in guix and gitian (MarcoFalke)
fa55a2554c depends: Remove reference to win32 (MarcoFalke)
Pull request description:
Closes#17504
Now that we no longer provide downloads for i686 on our website (https://bitcoincore.org/en/download/), there is no need to build them by default.
i686 can still be built in depends (tested by ci/travis) and in guix/gitian by setting the appropriate `HOSTS`.
ACKs for top commit:
practicalswift:
ACK fae9084ac5 -- patch looks correct
dongcarl:
ACK fae9084ac5 patch looks correct
laanwj:
Code review ACK fae9084ac5
hebasto:
ACK fae9084ac5, I have reviewed the code and it looks OK, I agree it can be merged.
Tree-SHA512: b000c19a2cd2a596a52028fa298c4022c24cfdfc1bdb3795a90916d0a00a32e4dd22278db93790b6a11724e08ea8451f4f05c77bc40d1664518e11a8c82d6e29
c491368d8c scripts: add MACHO dylib checking to symbol-check.py (fanquake)
76bf97213f scripts: fix check-symbols & check-security argument passing (fanquake)
Pull request description:
Based on #17857.
This adds dynamic library checks for MACHO executables to symbol-check.py. The script has been modified to function more like `security-check.py`. The error output is now also slightly different. i.e:
```bash
# Linux x86
bitcoin-cli: symbol operator new[](unsigned long) from unsupported version GLIBCXX_3.4
bitcoin-cli: export of symbol vtable for std::basic_ios<char, std::char_traits<char> > not allowed
bitcoin-cli: NEEDED library libstdc++.so.6 is not allowed
bitcoin-cli: failed IMPORTED_SYMBOLS EXPORTED_SYMBOLS LIBRARY_DEPENDENCIES
# RISCV (skips exported symbols checks)
bitcoin-tx: symbol operator new[](unsigned long) from unsupported version GLIBCXX_3.4
bitcoin-tx: NEEDED library libstdc++.so.6 is not allowed
bitcoin-tx: failed IMPORTED_SYMBOLS LIBRARY_DEPENDENCIES
# macOS
Checking macOS dynamic libraries...
libboost_filesystem.dylib is not in ALLOWED_LIBRARIES!
bitcoind: failed DYNAMIC_LIBRARIES
```
Compared to `v0.19.0.1` the macOS allowed dylibs has been slimmed down somewhat:
```diff
src/qt/bitcoin-qt:
/usr/lib/libSystem.B.dylib
-/System/Library/Frameworks/DiskArbitration.framework/Versions/A/DiskArbitration
/System/Library/Frameworks/IOKit.framework/Versions/A/IOKit
/System/Library/Frameworks/Foundation.framework/Versions/C/Foundation
/System/Library/Frameworks/CoreServices.framework/Versions/A/CoreServices
/System/Library/Frameworks/AppKit.framework/Versions/C/AppKit
/System/Library/Frameworks/ApplicationServices.framework/Versions/A/ApplicationServices
/System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation
-/System/Library/Frameworks/Security.framework/Versions/A/Security
-/System/Library/Frameworks/SystemConfiguration.framework/Versions/A/SystemConfiguration
/System/Library/Frameworks/CoreGraphics.framework/Versions/A/CoreGraphics
-/System/Library/Frameworks/OpenGL.framework/Versions/A/OpenGL
-/System/Library/Frameworks/AGL.framework/Versions/A/AGL
/System/Library/Frameworks/Carbon.framework/Versions/A/Carbon
/usr/lib/libc++.1.dylib
-/System/Library/Frameworks/CFNetwork.framework/Versions/A/CFNetwork
/System/Library/Frameworks/CoreText.framework/Versions/A/CoreText
/System/Library/Frameworks/ImageIO.framework/Versions/A/ImageIO
/usr/lib/libobjc.A.dylib
```
ACKs for top commit:
laanwj:
ACK c491368d8c
Tree-SHA512: f8624e4964e80b3e0d34e8d3cc33f3107938f3ef7a01c07828f09b902b5ea31a53c50f9be03576e1896ed832cf2c399e03a7943a4f537a1e1c705f3804aed979
fae75306ba scripted-diff: Set gitian arch back to amd64 (MarcoFalke)
Pull request description:
This was required to allow gitian builds on non-amd64 architecture, however, it seems to break the current builds (with lxc), see https://github.com/bitcoin/bitcoin/pull/17409#issuecomment-554099626
Also, the gititan builds wouldn't be deterministic across arches anyway, see #17468
So instead of wasting more time on this, revert the change and hope that guix allows to compile on non-amd64 architectures.
Top commit has no ACKs.
Tree-SHA512: 801e9a30ae1b0882ef45d5eb3a3cf80f3ace3b99db046069dbd95b6162119e977e3cf3134287d1ac5d09483906206acc71e1ac34d6b74dbc533d46aaf73f5cc2
feb5075777 build: use osslsigncode 2.0 in gitian (fanquake)
Pull request description:
The original osslsigncode project, https://sourceforge.net/projects/osslsigncode, has been marked as abandonware:
> This is now - and has been for a long while - abandonware. Feel free to create your own forks etc.".
However, a fork has emerged, https://github.com/mtrojnar/osslsigncode, that has incorporated
theuni's patches ([add the -pem option in extract-signature mode ](36715c1183) & [add the attach-signature command](3be7eb1676)) as well as updated the tool to work with OpenSSL 1.1 and other improvements.
This commit switches the windows signer descriptor to use this new version of `osslsigncode`.
I've tested using this new version of `osslsigncode` while doing a 0.18.1 gitian build, and it "seems" to work. However this needs a look over from Cory, to check if the tool is still compatible with his usage in the [`detached-sig-create.sh`](https://github.com/bitcoin/bitcoin/blob/master/contrib/windeploy/detached-sig-create.sh) script, as well as some review of the changes to `osslsigncode` itself. Hence WIP and chasing Concept ACKs / NACKs.
ACKs for top commit:
MarcoFalke:
Concept ACK feb5075777 given that this upstream is now used in Ubuntu and Debian
laanwj:
ACK feb5075777
Tree-SHA512: c48de6dc32751d96dd04b920bfacca40af47a2883330ba0700371d56c580a7e45cedd8d8a913709d56be036762b63cb1825a98cff7aa77b6d7804fab11220850
The original osslsigncode project (https://sourceforge.net/projects/osslsigncode/) has been marked as abandonware,
"This is now - and has been for a long while - abandonware. Feel free to create your own forks etc.".
However, a fork at https://github.com/mtrojnar/osslsigncode has emerged that has incorporated
theuni's patches, updated the tool to work with OpenSSL 1.1 and made other improvements.
This commit switches the windows signer descriptor to use this new version of osslsigncode.
bd3f5a90ec build: remove mingw linker workaround from win gitian descriptor (fanquake)
Pull request description:
This workaround was added as part of the switch to gitian building using Ubuntu 14.04 (#6900).
However, it should no longer be required, as we have switched to Bionic (#13171) and that
has a far newer version of binutils.
Original discussion: https://github.com/bitcoin/bitcoin/pull/6900
binutils patch: https://sourceware.org/bugzilla/show_bug.cgi?id=16192
ACKs for top commit:
MarcoFalke:
ACK bd3f5a90ec
theuni:
ACK bd3f5a90ec
laanwj:
ACK bd3f5a90ec
Tree-SHA512: 01a5789994decf8cdedf7aaa0a449d2100a77e2e6b422d6b9dd5a4ac3e2e0b538c3d43aae4a1c3713614782f3c6b09d8d8bb21c20e86ce3c1734183dedd02d0c
9d1f971c75 gitian: Put things in the right place to begin with (Carl Dong)
71949a97a7 gitian: Eliminate rename dependency (Carl Dong)
999a9a5f5b gitian: Smaller diff with gitian-linux.yml (Carl Dong)
c4a3c25ba1 gitian: Fix README inclusion in archives (Carl Dong)
93cb974980 gitian: Use split-debug.sh for Win builds (Carl Dong)
Pull request description:
It would seem that our `gitian-win.yml` has not been keeping up with `gitian-linux.yml`, this PR:
1. Minimizes the diff size between `gitian-{win,linux}.yml`
2. Eliminates the `rename` dependency
ACKs for top commit:
laanwj:
ACK 9d1f971c75
Tree-SHA512: 84ed47c685e12d0064c02811907ae3d0fd3c47db8773d497dcc38f0defbfb3040fd82899fb026cf355f229b906d05a1c8038a95642bb90d044afbc2e0b239af2
Linux:
The README was originally added in 8550f1fb2, but included the README
under the docs directory, which has a bunch of internal links that won't
make sense in a release tarball. In this patch, we include the root
level README instead, which makes more sense.
Windows:
.md files are inconvenient to open on windows and the line endings
differ, so we use README_windows.txt instead.
Debug splitting was first introduced in 7e7eb2724, then gitian-linux.yml
changed to using split-debug.sh in 9d2536208. Here we change
gitian-win.yml to use split-debug.sh as well.
This workaround was added as part of the switch to gitian building using Ubuntu 14.04 (#6900).
However, it should no longer be required, as we have switched to Bionic (#13171), and that
has a far newer version of binutils.
binutils patch: https://sourceware.org/bugzilla/show_bug.cgi?id=16192
32da92bdf6 gitian: Improve error handling (Wladimir J. van der Laan)
Pull request description:
Improve error handling in gitian builds:
- Set fail-on-error and pipefail flag, this causes a command to fail when either of the pipe stages fails, not only when the last of the stages fails, so this improves error detection.
- Also use `xargs` instead of `find -exec`, because `find` will not propagate errors in the executed command, but `xargs` will.
This will avoid some issues like #15541 where non-determinism is silently introduced due to errors caused by environment conditions (such as lack of disk space in that case).
Tree-SHA512: d5d3f22ce2d04a75e5c25e935744327c3adc704c2d303133f2918113573a564dff3d3243d5569a2b93ee7eb0e97f8e1b1ba81767e966af9015ea711a14091035
284f424d5a Fix osslsigncode compile issue in gitian-build (Chun Kuan Lee)
Pull request description:
Install libssl1.0-dev that is compatible with osslsigncode.
Fixes#13762
Verifed that this gitian descriptor file can sign 0.16.2rc2.
Tree-SHA512: 3029b86e77567a4e033b5ad95826e60df12a0486ac3c4afcac48218f5c76ba49e7f1c1307ce93ffc465ca2f24e12c401e4542929263688e4bd6521aeca3ff73b
6b5506a286 Fix Qt's rcc determinism for depends/gitian (Fuzzbawls)
Pull request description:
With the update to Qt 5.9 having been merged, Qt's `rcc` tool now embeds a file's last modified time in it's output. Since the build system generates temporary files for all locale translations (`*.qm` files) at build time, the resulting `qrc_bitcoin_locale.cpp` file was always being generated in a non-deterministic way.
This is a backport of https://bugreports.qt.io/browse/QTBUG-62511, which is included in Qt versions 5.11+, that allows for an environment variable (`QT_RCC_SOURCE_DATE_OVERRIDE`) to override the behavior described above. This environment variable is in turn set in the gitian descriptors, as that is where determinism is vital for release purposes.
Prior to this, the `qt_libbitcoinqt_a-qrc_bitcoin_locale.o` object file (included into `libbitcoinqt.a`) was returning a different `sha256sum` for each and every build, regardless of file contents change, thus breaking determinism in the resulting binaries.
This should fix#13731
Tree-SHA512: 174017e41f9afc3950ef54a9419de81577ec900db9aec3c78ccd3d879c6aecaaeb944fde0615b933f43e6ca9d7898a27ec071cdd0b91cb772755a3012de96725
Release version strings were broken in Gitian by 7522. This is a minimal fix
suitable for 0.15.
After this, we should fix up version handling for good so that gitian packages
the correct string in the release tarball, so that git is not required to get
the tag name.
1d8df0141 Fix MD formatting in REST-interface.md and spelling mistake in test_runner.py (MeshCollider)
41f3e84aa Fix inconsistencies and grammar in various files (MeshCollider)
Pull request description:
Just a simple fix of some inconsistent capitalization, formatting and grammar in a few files (no code changes)
Tree-SHA512: 60b12a5a5c69a1af4a25b7db0b32ed806ed62ad2966cee08b3792a7cfa7f51848fd485349b4c09e60a7eedfdf55ee730c51daa066d6e226ae404c93342bf3e13
ed1fcdc Bugfix: Detect genbuild.sh in repo correctly (Luke Dashjr)
e98e3dd Bugfix: Only use git for build info if the repository is actually the right one (Luke Dashjr)
Tree-SHA512: 510d7ec8cfeff4e8e0c7ac53631eb32c7acaada7017e7f8cc2e6f60d86afe1cd131870582e01022f961c85a783a130bcb8fef971f8b110070c9c02afda020726
09fe2d9 release: update docs to show basic codesigning procedure (Cory Fields)
f642753 release: create a bundle for the new signing script (Cory Fields)
0068361 release: add win detached sig creator and our cert chain (Cory Fields)
Tree-SHA512: 032ad84697c70faaf857b9187f548282722cffca95d658e36413dc048ff02d9183253373254ffcc1158afb71140753f35abfc9fc8781ea5329c04d13c98759c0
Also change the mac filename to match
The procedure remains the same, but now there's a nifty script to automate
the signing process.
Future steps:
- Build osslsigncode in the gitian-win descriptor so that the signer itself is
deterministic.
- Verify in the gitian-win-signer descriptor that the expected cert chain was
used.
The consistency is helpful for gauging Gitian build progress. Right now it's necessary to remember which platform builds in which order, which can be confusing if you're attempting to get a quick idea of how far along your builds are.
C_INCLUDE_PATH and CPLUS_INCLUDE_PATH work globally as though -isystem was used
for each invocation.
Since that changes the build results, force a rebuild of x86 depends by adding
the value to $HOST_ID_SALT.
- create a script to handle split debug. This will also eventually need to check
targets, and use dsymutil for osx.
- update config.guess/config.sub for bdb for aarch64.
- temporarily disable symbol checks for arm/aarch64
- quit renaming to linux32/linux64 and use the host directly
This also adds a hack to work around an Ubuntu bug in the gcc-multilib package:
https://bugs.launchpad.net/ubuntu/+source/gcc-defaults-armhf-cross/+bug/1347820
The problem is that gcc-multilib conflicts with the aarch toolchain.
gcc-multilib installs a symlink that points
/usr/include/asm -> /usr/include/x86_64-linux-gnu/asm.
Without this link, gcc -m32 can't find asm/errno.h (and others), since
/usr/include/x86_64-linux-gnu isn't in its default include path. But
/usr/include/i386-linux-gnu is (though it doesn't exist on disk).
So work around the problem by linking
/usr/include/i386-linux-gnu/asm -> /usr/include/x86_64-linux-gnu/asm.
The symlink fix is actually quite reasonable, but echoing the password into
sudo is nasty, and should probably be addressed in gitian itself. It makes more
sense to enable passwordless sudo for the build user by default.
The -debug tarballs/zips contain detached debugging symbols. To use them, place
in the same dir as the target binary, and invoke gdb as usual.
Also, because the debug symbols add a substantial space requirement, the build
dirs are now deleted when they're no longer needed.
This removes the following executables from the binary gitian release:
- test_bitcoin-qt[.exe]
- bench_bitcoin[.exe]
@jonasschnelli and me discussed this on IRC a few days ago - unlike the
normal `bitcoin_tests` which is useful to see if it is safe to run
bitcoin on a certain OS/environment combination, there is no good reason
to include these. Better to leave them out to reduce the download
size.
Sizes from the 0.12 release:
```
2.4M bitcoin-0.12.0/bin/bench_bitcoin.exe
22M bitcoin-0.12.0/bin/test_bitcoin-qt.exe
```
Common sentiment is that the miniupnpc codebase likely contains further
vulnerabilities.
I'd prefer to get rid of the dependency completely, but a compromise for
now is to at least disable it by default.
fanquake