33796a964a build: Add ability to build qt in depends with -stdlib=libc++ (Hennadii Stepanov)
Pull request description:
This PR makes possible to build the `qt` package in depends against `libc++` for x86_64 platform.
Fixes#22344.
Required for #22815.
Also this PR [fixes](https://github.com/bitcoin/bitcoin/pull/23060#discussion_r716077050) the `[no wallet] [bionic]` task on CI:
- on master (a8bbd4cc81), https://api.cirrus-ci.com/v1/task/5558609250615296/logs/ci.log:
```
Options used to compile and link:
external signer = yes
multiprocess = no
with libs = yes
with wallet = no
with gui / qt = no
```
- this PR, https://api.cirrus-ci.com/v1/task/5502605561430016/logs/ci.log:
```
Options used to compile and link:
external signer = yes
multiprocess = no
with libs = yes
with wallet = no
with gui / qt = yes
```
ACKs for top commit:
fanquake:
ACK 33796a964a - While this sort of string matching is fragile, I think the risk of this causing any actual issues is low.
Tree-SHA512: 586dde2e9864cec7a49aeb4f2b77fb8c4ae96bd10b51f9c6de0cfe8512ad61db15bb7f8d1b0eb6a5a66fd2deee52ac52218f01eb6be107ac12f1a956190de54b
d3b0f82a43 build: Fix regression introduced in PR23603 (Hennadii Stepanov)
Pull request description:
It appears 7629efcc2c from bitcoin/bitcoin#23603 introduced a regression in build tool flag evaluation.
On macOS system:
- pre-PR23603 master (ae017b8160):
```
% make -C depends print-darwin_CXXFLAGS
darwin_CXXFLAGS=-pipe
% make -C depends print-host_CXXFLAGS
host_CXXFLAGS=-pipe
```
- the current master (369978686e):
```
% make -C depends print-darwin_CXXFLAGS
darwin_CXXFLAGS=-pipe
% make -C depends print-host_CXXFLAGS
host_CXXFLAGS=-arch x86_64
```
It's obvious a flag being set in `depends/hosts/darwin.mk`, i.e., `-pipe`, is lost.
With this PR:
```
% make -C depends print-darwin_CXXFLAGS
darwin_CXXFLAGS=-pipe
% make -C depends print-host_CXXFLAGS
host_CXXFLAGS=-pipe -arch x86_64
```
ACKs for top commit:
fanquake:
ACK d3b0f82a43
Tree-SHA512: 643099ce6858475ac9f3a4dfa72a4e493fec6fdd7042ae0f0d5fe44c5cd175e4eda63cb39fc46ac1501cadcd3466507ec88d9089235e005fe43ea7ab47ce37c1
11736dbe3d build, qt: Hardcode last modified timestamp in Qt RCC (Hennadii Stepanov)
Pull request description:
Our Guix build system sets the [`SOURCE_DATE_EPOCH`](https://reproducible-builds.org/specs/source-date-epoch/) and propagates it to the depends build subsystem. Its [default value](https://github.com/bitcoin/bitcoin/blob/master/contrib/guix/README.md#recognized-environment-variables) is the top commit timestamp.
After bumping Qt version up to 5.15.2, due to [this](1ffcca4cc2) change, every time they are going to make new Guix builds for another branch/commit they must ensure that the `qt` package will be rebuilt from scratch. Otherwise, Bitcoin Core GUI binaries will be non-deterministic.
Such behavior makes working with Guix builds suboptimal.
This PR fixes the described issue by patching Qt RCC and hardcoding last modified timestamps with `1`.
It's worth to mention that this change is compatible with a possible future [improvement](https://github.com/bitcoin/bitcoin/pull/21995) which makes each dependency package reproducible.
A drawback of such an approach is not currently applied to our project, as it effectively makes [QML cache files](https://bugreports.qt.io/browse/QTBUG-57182) useless. I can't say it's a problem for the https://github.com/bitcoin-core/gui-qml project.
---
**A note for thinkers:** For now this change is enough as only Qt source contains `SOURCE_DATE_EPOCH`. But in general we should re-think about treating the `SOURCE_DATE_EPOCH` variable in the depends build subsystem. For instance, its default value could be the output of `git log --format=%at -1 -- depends`.
ACKs for top commit:
fanquake:
ACK 11736dbe3d
Tree-SHA512: 31f104010a0a78d217aafcc5bc4606351f9060fc2a827277935b85fc8ced9f3d90a31d812c7db8c2711fb6daccd279cf0945dc1d7a7199e0eb0ade451cdbcd5d
8494dcae0e Replace $(AT) with .SILENCE. (Dmitry Goncharov)
Pull request description:
This reduces the amount of syntax noise in the makefiles.
Setting V=1 still enables verbose logging.
The only noticeable difference in behavior is that, unless V=1 is specified, make won't print its own messages like
make: Nothing to be done for 'all', make: 'all' is up to date, or touch <file>, if -t is specified.
ACKs for top commit:
laanwj:
Tested ACK 8494dcae0e
Tree-SHA512: 66b9111229995aa54a9e87f4571648727d89b8529caec651063cdfe5c00a64341371b648701d192b2334df0614617a00c28eaa56c7f08ee9c00127cada0293ab
66a20a54a2 build, qt: Drop support for `i686-linux-android` host (Hennadii Stepanov)
Pull request description:
There are no reasons to keep support for `i686-linux-android` host, which is actually broken in master (50c502f54a), and this fact has been unnoticed for months :)
https://github.com/bitcoin/bitcoin/pull/23675#issuecomment-986206434:
> I'm surprised `i686-linux-android` ABI is still supported. I would love to drop it...
https://github.com/bitcoin/bitcoin/pull/23675#issuecomment-991340132
> What is `i686-linux-android`? 32-bit x86 android? is that really a thing?
ACKs for top commit:
prusnak:
utACK 66a20a54a2
Tree-SHA512: 211f794de2fc569f0ade2a4da805b8bfd4ce2ab0930c5d427acc4f5d015fcdc4911f02fc64f6401197f7641aed79944a9594be80c817547be3269cdd721cf79b
Since Qt 5.14.0 the CROSS_COMPILE variable always being overridden by
the Qt build system internally.
See upstream commit: c28b881c98fadcd3415370fad2525b558f6b03e4
19f3896c9a build: Fix x86_64 <-> arm64 cross-compiling for macOS (Hennadii Stepanov)
Pull request description:
Currently, on master (111c3e06b3), dependencies are built for the build system architecture, not the provided host.
On Intel-based macOS Big Sur 11.6.1 (20G224):
```
% make -C depends HOST=arm64-apple-darwin20
% lipo -info depends/arm64-apple-darwin20/lib/libsqlite3.a
Non-fat file: depends/arm64-apple-darwin20/lib/libsqlite3.a is architecture: x86_64
```
On M1-based macOS Monterey 12.0.1 (21A559) the `boost` package building fails with multiple errors like that:
```
% make -C depends boost HOST=x86_64-apple-darwin19
...
error: option 'cf-protection=return' cannot be specified on this target
error: option 'cf-protection=branch' cannot be specified on this target
2 errors generated.
```
This PR allows to cross-compile as follows:
- on Intel-based macOS Big Sur 11.6.1 (20G224):
```
% make -C depends HOST=arm64-apple-darwin20
% lipo -info depends/arm64-apple-darwin20/lib/libsqlite3.a
Non-fat file: depends/arm64-apple-darwin20/lib/libsqlite3.a is architecture: arm64
% CONFIG_SITE=$PWD/depends/arm64-apple-darwin20/share/config.site ./configure
% make
% lipo -info src/qt/bitcoin-qt
Non-fat file: src/qt/bitcoin-qt is architecture: arm64
```
- on M1-based macOS Monterey 12.0.1 (21A559):
```
% make -C depends HOST=x86_64-apple-darwin19
% CONFIG_SITE=$PWD/depends/x86_64-apple-darwin19/share/config.site ./configure
% make
% lipo -info src/qt/bitcoin-qt
Non-fat file: src/qt/bitcoin-qt is architecture: x86_64
```
No behavior change for other builder-host pairs.
This is an alternative to bitcoin/bitcoin#22506.
ACKs for top commit:
fanquake:
ACK 19f3896c9a - this is definitely cleaner than the approach in #22506. I've tested the x86_64 -> arm64 and arm64 -> x86_64 cross-compiles. Going to go-ahead and merge this given it's scoped to building for darwin targets, on darwin. Not sure if this is something we'll actually see used very often.
Tree-SHA512: 67a7fed1db79bc43d41aabb5a676f90417a879e5dcffec1384c288b56ee32ba06ae3819abdc7aca85b81af6e782bcfff5c83a67a455560249fa095a807fc1187
6575d354c8 build: Bump Fonconfig version up to 2.12.6 (Hennadii Stepanov)
Pull request description:
This PR gets rid of `remove_char_width_usage.patch`.
Some additional observations:
1. Newer Fontconfig versions (2.13.0 and 2.13.1) introduce a new dependency, `uuid`, in the [`7b48fd3dd406b926f0e5240b211f72197ed538a9`](7b48fd3dd4) commit
2. In Fonconfig 2.13.1 (the current stable) excludes the `fcobjshash.h` from the distributive archive (see [`31269e3589e0e6432d12f55db316f4c720a090b5`](31269e3589)), that makes our `gperf_header_regen.patch` unusable, and requires `gperf` as a dependency.
ACKs for top commit:
fanquake:
ACK 6575d354c8 - from the best I can determine this doesn't have any versioning / ABI implications. The ABI difference between 2.12.1 and 2.12.6 is two symbol additions, neither of which are used by Qt. Fontconfig seems to be better at maintaining backwards compatibility compared to a library like Freetype.
Tree-SHA512: 36780a0c5a658469697e524d682ebab56c320cb04f8297bc215f4552f183d4f560501fb0a869982fd9053d4a2d571c7fd971d8f5e96c9da9a9d142c485e3baa4
2f356a0ca8 scripted-diff: Drop Darwin version for better maintainability (Hennadii Stepanov)
Pull request description:
After this PR, any macOS tools version bumping in the future will touch fewer files in the repo.
Pointing a Darwin version for the `--host` system does not matter for the following reasons:
- in terms of the resulted binaries, we should only care about the minimum supported macOS version which is a separated parameter in our build system.
- in terms of the build system itself, the usage of the `$(host)` variable is self-consistent enough. Btw `$(host_os)` value already has the version dropped:
```
$ make -C depends --no-print-directory print-host_os HOST=x86_64-apple-darwin19
host_os=darwin
```
ACKs for top commit:
gruve-p:
ACK 2f356a0ca8
promag:
ACK 2f356a0ca8.
fanquake:
ACK 2f356a0ca8
Tree-SHA512: 374896ab0ba02b0d8b4b21431fe963bd213b0d09586e0898c13a4c5fa294c1b693f1b2c92880c245c4157c14217b4825b36522f461930477f4d2a727086ebb2a
78a6bc6919 build, qt: Use Android NDK r23 LTS (Hennadii Stepanov)
Pull request description:
This is a continuation of bitcoin/bitcoin#23478, and, thanks to bitcoin/bitcoin#23489, a oneline patch is only required to be able build the `qt` package in depends with Android NDK r23 LTS.
ACKs for top commit:
fanquake:
ACK 78a6bc6919
Tree-SHA512: 09c6e8739ecbcbf5fdd6c2103577be2676eb448941f97c781f476918056c8405d2531d5cef8f240e4d1205c2d49f879edbba74dd5e77799a887b76a5c76ebe5b
Very old shells suffered from bugs which meant that prefixing variables
with an "x" to ensure that the lefthand side of a comparison always
started with an alphanumeric character was needed. Modern shells don't
suffer from this issue (i.e Bash was fixed in 1996).
In any case, we've already got unprefixed checks used in our codebase,
i.e https://github.com/bitcoin/bitcoin/blob/master/configure.ac#L292,
and have dependencies (in depends) that also use unprefixed comparisons.
I think it's time that we can consolidate on not using the x-prefix
workaround. At best it's mostly just confusing.
More info:
https://github.com/koalaman/shellcheck/wiki/SC2268https://www.vidarholen.net/contents/blog/?p=1035
855bd75503 build, qt: Ditch `no_sdk_version_check.patch` (Hennadii Stepanov)
Pull request description:
Since Qt 5.12.2 (see 5d181961ea) it is possible to control whether to check macOS SDK with the `QT_MAC_SDK_NO_VERSION_CHECK` variable.
ACKs for top commit:
fanquake:
ACK 855bd75503
Tree-SHA512: e490f22da9c37d25b58418ca62e65d2303cfbd83139f4abea4b67c37fa97e9a882ccc9d484b49cd2bb3a8851ea35e63123d3c6db7f804d6b50f41b10bd10c10b
4ba492052e doc: Add minimum supported Android NDK version (Hennadii Stepanov)
6393bdcd53 doc: Move Android dependencies guide into `build-android.md` (Hennadii Stepanov)
ac323a7222 build: Switch to llvm buinutils for Android builds (Hennadii Stepanov)
Pull request description:
The new Long Term Support release of the Android NDK is [available](https://groups.google.com/g/android-ndk-announce/c/MS6Qoub0DKE/m/Zfp5Ys8eAAAJ) since 2021-08-11:
> As r23 is the new LTS, the support windows for r21 and r22 have now ended.
On master (8ae4ba481c), dependency build fails because it expects GNU Binutils are present in the Android NDK. In [fact](https://android.googlesource.com/platform/ndk/+/master/docs/BuildSystemMaintainers.md#binutils):
> GNU Binutils remains available up to and including r22. All binutils tools with the exception of the assembler (GAS) were removed in r23. GAS was removed in r24.
This PR switches our depends build system to llvm binutils. The usage of `llvm-ar` and `llvm-ranlib` tools effectively makes r21 the minimum supported version of NDK.
With this PR:
- building depends against NDK r23 LTS now is possible with `NO_QT=1`
- building the `qt` package in depends against NDK r23 LTS still fails:
```
Creating qmake...
...
ERROR: Cannot detect Android NDK toolchain.
Please use -android-toolchain-version to specify it.
```
The issue with the `qt` package is going to be addressed in another PR.
ACKs for top commit:
fanquake:
ACK 4ba492052e
Tree-SHA512: cdc8f95ff9a3ad7f12eb55b9ea18b6b6b800d4cceff7e0321985be6e39d15a2b2ea5b1592972307d76d111292a0ed58fd287e5ca285e2f6868b42a286536d310
