Without ffile-prefix-map, the debug symbols will contain paths for the
guix store which will include the hashes of each package. However, the
hash for the same package will differ when on different architectures.
In order to be reproducible regardless of the architecture used to build
the package, map all guix store prefixes to something fixed, e.g. /usr.
Now that we use GCC 10 for release builds, we no-longer need to
pass-Wl,-z,noexecstack to get a non-executable stack in RISC-V binaries.
This was originally removed in #21036, but then re-added in #21799, when
we reverted to using GCC 8.
copy over bitcoin.conf during the build process.
this means `contrib/devtools/gen-bitcoin-conf.sh` will need
to be run and the generated file committed during the release process.
this is the same process used for generating man pages for each release.
From what I can see the only platform this drops support for is CentOS
7. CentOS 7 reached the end of it's "full update" support at the end of
2020. It does receive maintenance updates until 2024, however I don't
think supporting glibc 2.17 until 2024 is realistic. Note that anyone
wanting to self-compile and target a glibc 2.17 runtime could build with
--disable-threadlocal.
glibc 2.18 was released in August 2013.
https://sourceware.org/legacy-ml/libc-alpha/2013-08/msg00160.html
0f95247246 Integrate univalue into our buildsystem (Cory Fields)
9b49ed656f Squashed 'src/univalue/' changes from 98fadc0909..a44caf65fe (fanquake)
Pull request description:
This PR more tightly integrates building Univalue into our build system. This follows the same approach we use for [LevelDB](https://github.com/bitcoin-core/leveldb/), ([`Makefile.leveldb.include`](https://github.com/bitcoin/bitcoin/blob/master/src/Makefile.leveldb.include)), and [CRC32C](https://github.com/bitcoin-core/crc32c) ([`Makefile.crc32c.include`](https://github.com/bitcoin/bitcoin/blob/master/src/Makefile.crc32c.include)), and will be the same approach we use for [minisketch](https://github.com/sipa/minisketch); see #23114.
This approach yields a number of benefits, including:
* Faster configuration due to one less subconfigure being run during `./configure` i.e 22s with this PR vs 26s
* Faster autoconf i.e 13s with this PR vs 17s
* Improved caching
* No more issues with compiler flags i.e https://github.com/bitcoin/bitcoin/pull/12467
* More direct control means we can build exactly the objects we want
There might be one argument against making this change, which is that builders should have the option to use "proper shared/system libraries". However, I think that falls down for a few reasons. The first being that we already don't support building with a number of system libraries (secp256k1, leveldb, crc32c); some for good reason. Univalue is really the odd one out at the moment.
Note that the only fork of Core I'm aware of, that actively patches in support for using system libs, also explicitly marks them as ["DANGEROUS"](a886811721/configure.ac (L1430)) and ["NOT SUPPORTED"](a886811721/configure.ac (L1312)). So it would seem they exist more to satisfy a distro requirement, as opposed to something that anyone should, or would actually use in practice.
PRs like #22412 highlight the "issue" with us operating with our own Univalue fork, where we actively fix bugs, and make improvements, when upstream (https://github.com/jgarzik/univalue) may not be taking those improvements, and by all accounts, is not currently actively maintained. Bitcoin Core should not be hamstrung into not being able to fix bugs in a library, and/or have to litter our source with "workarounds", i.e #22412, for bugs we've already fixed, based on the fact that an upstream project is not actively being maintained. Allowing builders to use system libs is really only exacerbating this problem, with little benefit to our project. Bitcoin Core is not quite like your average piece of distro packaged software.
There is the potential for us to give the same treatment to libsecp256k1, however it seems doing that is currently less straightforward.
ACKs for top commit:
dongcarl:
ACK 0f95247246 less my comment above, always nice to have an include-able `sources.mk` which makes integration easier.
theuni:
ACK 0f95247246. Thanks fanquake for keeping this going.
Tree-SHA512: a7f2e41ee7cba06ae72388638e86b264eca1b9a8b81c15d1d7b45df960c88c3b91578b4ade020f8cc61d75cf8d16914575f9a78fa4cef9c12be63504ed804b99
I used Guix's values for the powerpc64(le) dynamic linkers, and the
/lib-prefix seems to be a Guix-ism rather than standard. The standard
path for the linker-loaders start with /lib64.
I've taken the new loader values from SYSDEP_KNOWN_INTERPRETER_NAMES in
glibc's sysdeps/unix/sysv/linux/powerpc/ldconfig.h file.
For future reference, loader path values can also be found on glibc's
website: https://sourceware.org/glibc/wiki/ABIList?action=recall&rev=16
This addresses issues like the one in #12467, where some of our compiler flags
end up being dropped during the subconfigure of Univalue. Specifically, we're
still using the compiler-default c++ version rather than forcing c++17.
We can drop the need subconfigure completely in favor of a tighter build
integration, where the sources are listed separately from the build recipes,
so that they may be included directly by upstream projects. This is
similar to the way leveldb build integration works in Core.
Core benefits of this approach include:
- Better caching (for ex. ccache and autoconf)
- No need for a slow subconfigure
- Faster autoconf
- No more missing compile flags
- Compile only the objects needed
There are no benefits to Univalue itself that I can think of. These changes
should be a no-op there, and to downstreams as well until they take advantage
of the new sources.mk.
This also removes the option to use an external univalue to avoid similar ABI
issues with mystery binaries.
Co-authored-by: fanquake <fanquake@gmail.com>
1edddf5de4 Avoid GCC 7.1 ABI change warning in guix build (Pieter Wuille)
Pull request description:
The arm-linux-gnueabihf guix build output is littered with warnings like:
```
/gnu/store/7a96hdqdb2qi8a39f09n84xjy2hr23rs-gcc-cross-arm-linux-gnueabihf-8.4.0/include/c++/bits/stl_vector.h:1085:4: note:
parameter passing for argument of type '__gnu_cxx::__normal_iterator<CRecipient*, std::vector<CRecipient> >' changed in GCC 7.1
```
These are irrelevant for us. Disable them using `-Wno-psabi`.
ACKs for top commit:
laanwj:
ACK 1edddf5de4
hebasto:
ACK 1edddf5de4, after thorough reading related materials, I agree this change can be merged. As I mentioned above, I have been compiling my arm-32bit binaries with `-Wno-psabi` flag for two years, and no related flaws were observed.
Tree-SHA512: 485c7500547ac5da567ad23847341c18ff832607f5a1002676404cc647e437cf3445b6894ecff5b52929ca52bea946c06bd90eace1997c895e56204e787065e4
Now that our Guix builds are performed on glibc 2.24 and 2.27 (RISCV),
we no-longer need to pass the --enable-glibc-back-compat option.
Replace it with --disable-threadlocal, to prevent the usage of symbols
from glibc 2.18.
None of the binaries produced required symbols later than 2.17, and 2.27
(RISCV).
Our 'bitcoin-linux-g++' definition better integrates with our depends
system than the stock linux-g++-64 definition.
This fixes a bug whereby Guix builds on x86_64 for x86_64 did not
produce a QMinimalIntegrationPlugin and led to bitcoin-qt not being
built.
108a6be92a guix: Check for disk space availability before building (Carl Dong)
d7dec89091 guix: Remove dest if OUTDIR mv fails (Carl Dong)
Pull request description:
There seems to be some corner cases that can be hit when guix scripts unexpectedly fail in the middle of operation, see: https://gnusha.org/bitcoin-builds/2021-05-24.log
- Perform an early disk space check for `guix-build`
- Overwrite existing output directory after a successful build (the existing one might be malformed), and cleanup output directory if the `mv` somehow fails
ACKs for top commit:
laanwj:
Tested ACK 108a6be92a
achow101:
ACK 108a6be92a
Tree-SHA512: cf6438317da40bf55714cd2d8cce859b3d435cc66cabefe8d4a53552d7880966acfe84ffe8fadf1c80e368ae6b037992258a6d409df85ffc6ce8bf780e98e2e5
a58868d201 build: Makes rcc output always deterministic (Hennadii Stepanov)
Pull request description:
The Qt Resource Compiler ([rcc](https://doc.qt.io/qt-5/rcc.html)) has a command-line option `--format-version` which has the [default value](https://code.qt.io/cgit/qt/qtbase.git/tree/src/tools/rcc/main.cpp?h=5.12.10#n172) 2.
The only difference from `--format-version 1` is adding a [last modified timestamp](https://code.qt.io/cgit/qt/qtbase.git/tree/src/tools/rcc/rcc.cpp?h=5.12.10#n207) to the output file ([credits](https://github.com/bitcoin/bitcoin/pull/21654#issuecomment-819198228) to **fanquake**). That, in turn, forces us to use `QT_RCC_SOURCE_DATE_OVERRIDE=1` to get deterministic builds (#13732).
This change makes rcc output always deterministic by using `--format-version 1` option that makes usage of the
`QT_RCC_SOURCE_DATE_OVERRIDE` needless.
---
Also it improves interaction with ccache:
On master (f6c44e999b):
```
$ make && make clean && ccache --zero-stats && make && ccache --show-stats
...
cache directory /home/hebasto/.ccache
primary config /home/hebasto/.ccache/ccache.conf
secondary config (readonly) /etc/ccache.conf
stats updated Sun Apr 11 15:45:43 2021
stats zeroed Sun Apr 11 15:45:05 2021
cache hit (direct) 638
cache hit (preprocessed) 0
cache miss 1
cache hit rate 99.84 %
called for link 10
cleanups performed 0
files in cache 20023
cache size 13.2 GB
max cache size 15.0 GB
```
The missed file is always `qt/libbitcoinqt_a-qrc_bitcoin_locale.o`.
With this PR:
```
$ make && make clean && ccache --zero-stats && make && ccache --show-stats
...
cache directory /home/hebasto/.ccache
primary config /home/hebasto/.ccache/ccache.conf
secondary config (readonly) /etc/ccache.conf
stats updated Sun Apr 11 15:28:46 2021
stats zeroed Sun Apr 11 15:28:21 2021
cache hit (direct) 639
cache hit (preprocessed) 0
cache miss 0
cache hit rate 100.00 %
called for link 10
cleanups performed 0
files in cache 20012
cache size 13.2 GB
max cache size 15.0 GB
```
ACKs for top commit:
fanquake:
ACK a58868d201
Tree-SHA512: 52f4a3267f41883d13025c0de79b6da22e92d60c729e01b986935c6812bbfe7fadc40b742bd715bfdf09df94af6838d4fbbe8208c6123f366108e38c8e1121c5
d420e5c1c0 guix-attest: Avoid incomplete sigdirs with ERR traps (Carl Dong)
feda2c8e31 guix: Skip attesting to dist-archive (Carl Dong)
d522d8006b guix: Attest to inputs in inputs.SHA256SUMS (Carl Dong)
f9e2960c01 guix: Construct $OUTDIR in ${DISTSRC}/output (Carl Dong)
022abc85fc guix: Minor quoting fix in libexec/build.sh (Carl Dong)
c83c4fa5b7 guix-attest: Allow skipping GPG signing with NO_SIGN (Carl Dong)
0e1c2e448c guix-attest: Use ascii-armor signatures (Carl Dong)
b5fd89c4c8 guix-attest: Only use cross-platform flags for find+xargs (Carl Dong)
5926432ba6 guix: Add guix-verify script (Carl Dong)
30daf76a97 guix: Add guix-attest script (Carl Dong)
Pull request description:
Adds replacements for `gsign` and `gverify`.
Personally I'm not a big fan of using the word "sign" as it's been used to refer to both codesigning and GPG signing.
ACKs for top commit:
laanwj:
Code review and tested ACK d420e5c1c0
Tree-SHA512: 93d82d201f4596eaea0e3825aa55b013dfb91790e6ccee79893833d37921513d7b4e735f0641103e1e2ea8308abe4cb6218b73160924708802f2e0e3f7f6caf1
The Qt Resource Compiler (rcc) has a command-line option
`--format-version` which has the default value 2.
The only difference from `--format-version 1` is adding a last modified
timestamp to the output file. That, in turn, forces us to use
`QT_RCC_SOURCE_DATE_OVERRIDE=1` to get deterministic builds.
This change makes rcc output always deterministic by using
`--format-version 1` option that makes usage of the
`QT_RCC_SOURCE_DATE_OVERRIDE` needless. Also it improves interaction
with ccache.
Co-authored-by: fanquake <fanquake@gmail.com>
46b025e00d test: add new python linter to check file names and permissions (windsok)
6f6bb3ebc7 test: fix file permissions on various scripts (windsok)
Pull request description:
Adds a new python linter test which tests for correct filenames and file permissions in the repository.
Replaces the existing tests in the `test/lint/lint-filenames.sh` and `test/lint/lint-shebang.sh` linter tests, as well as adding some new and increased testing. This increased coverage is intended to catch issues such as in #21728 and https://github.com/bitcoin/bitcoin/pull/16807/files#r345547050
Summary of tests:
* Checks every file in the repository against an allowed regexp to make sure only lowercase or uppercase alphanumerics (a-zA-Z0-9), underscores (_), hyphens (-), at (@) and dots (.) are used in repository filenames.
* Checks only source files (*.cpp, *.h, *.py, *.sh) against a stricter allowed regexp to make sure only lowercase alphanumerics (a-z0-9), underscores (_), hyphens (-) and dots (.) are used in source code filenames. Additionally there is an exception regexp for directories or files which are excepted from matching this regexp (This should replicate the existing `test/lint/lint-filenames.sh` test)
* Checks all files in the repository match an allowed executable or non-executable file permission octal. Additionally checks that for executable files, the file contains a shebang line.
* Checks that for executable `.py` and `.sh` files, the shebang line used matches an allowable list of shebangs (This should replicate the existing `test/lint/lint-shebang.sh` test)
* Checks every file that contains a shebang line to ensure it has an executable permission
Additionally updates the permissions on various files to comply with the new tests.
Fixes#21729
ACKs for top commit:
practicalswift:
cr re-ACK 46b025e00df40724175735eb5606ac73067cb3b8: patch still looks correct
kiminuo:
code review ACK 46b025e00d if `contrib/gitian-descriptors/assign_DISTNAME` permission change is deemed OK.
laanwj:
Code review ACK 46b025e00d
Tree-SHA512: 1c8201a2cee0d9cbce15652b68cec9a6458a8b493fcd5392f98560aca0b1a12e668baab65a47100f116f626dadc3f591deb47f7368468c6a46c6c712c2533455
We already attest to the relevant dist-archive in inputs.SHA256SUMS,
which is recorded at build-time.
We use a SKIPATTEST.TAG file to indicate output directories which do not
require attestation (much like the CACHEDIR.TAG specification).
Generally, it's better to have build scripts declare properties of
directories instead of introducing name-based special cases in attest
scripts since build scripts have a more detailed context of what is
going on.
At build/codesigning-time, hash build inputs and output the digest to
${OUTDIR}/inputs.SHA256SUMS, which gets included in the final SHA256SUMS
constructed by guix-attest.
Example final SHA256SUMS:
ee832d2a35b7701bff581dea05a536118b118e3ad0a587a2855b6ee8cd6fba20 inputs/bitcoin-78199266af7b.tar.gz
ca765e70a0c12866dd63c0be228b675278a26329e5f8f5b5c52fd09200fedf21 bitcoin-78199266af7b-powerpc64le-linux-gnu-debug.tar.gz
dae95327d7f2c324e2728c4b73627be6cb2c0d2f2e5bea940d1d5e6463939327 bitcoin-78199266af7b-powerpc64le-linux-gnu.tar.gz
While files are being output to $OUTDIR, it will be under
${DISTSRC}/output, and only when everything is done, will
${DISTSRC}/output be moved to the actual $OUTDIR.
This makes it so that a Ctrl-C in the middle of a build is less likely
to result in a partially-constructed $OUTDIR. In fact, if I understand
correctly, if $OUTDIR and $DISTSRC reside on the same filesystem, the
move (rename) is likely atomic.
Also, since the "working $OUTDIR" is under ${DISTSRC}/output, it will be
cleaned properly by the guix-clean script.
This relatively easy change eliminates all runtime dependencies (except
for the kernel) for dmg, which is the only native build tool that gets
put in our output tarballs.
This allows much more flexibility when constructing the codesigning
environment, and is much more robust.
./windeploy is a "working directory", and therefore belongs inside
distsrc-*. Many people have noticed their Guix builds failing after
hours simply because they did not remove windeploy (but did remove the
distsrc-* directories).
In Guix, there are two flags for controlling parallelism:
Note: When I say "derivation," think "package"
--cores=n
- controls the number of CPU cores to build each derivation. This is
the value passed to `make`'s `--jobs=` flag.
- defaults to 0: as many cores as is available
--max-jobs=n
- controls how many derivations can be built in parallel
- defaults to 1
Therefore, if set --max-jobs=$MAX_JOBS and don't set --cores, Guix could
theoretically spin up $MAX_JOBS * $(nproc) number of threads, and that's
no good.
So we could either default to --cores=1, --max-jobs=$MAX_JOBS
- Pro: --cores=1 means that `make` will be invoked with `-j1`,
avoiding problems with package whose build systems and test
suites break when running multi-threaded.
- Con: There will be times when only 1 or 2 derivations can be built
at a time, because the rest of the dependency graph all depend
on those 1 or 2 derivations. During these times, the machine
will be severely under-utilized.
or --cores=$MAX_JOBS, --max-jobs=1
- Pro: We don't encounter prolonged periods of
severe under-utilization mentioned above.
- Con: Many packages' build systems and test suites break when running
multi-threaded.
or --cores=1, --max-jobs=1 and let the user override with
$ADDITIONAL_GUIX_COMMON_FLAGS
13a9fd11a5 guix: Passthrough SDK_PATH into container (Carl Dong)
Pull request description:
This is a usability improvement for Guix builders so that they don't have to extract the Xcode tarball into `depends/SDKs` every time.
Inspiration: https://github.com/bitcoin/bitcoin/pull/21089#issuecomment-778639698
ACKs for top commit:
laanwj:
Tested ACK 13a9fd11a5
Tree-SHA512: 63392d537e48a0da9f0ee04a929613b139bef1ac5643187871c9ea5376afd2a3d95df0f5e0950ae0eccd2813b166667be98401e5a248ae9c187fe4e84e54d427
Hennadii Stepanov