Compiling fails on 32-bit systems

hebasto commented at 1:31 pm on September 4, 2021: member

Since #20233 compiling with clang 10.0 fails:

0bench/addrman.cpp:113:32: error: non-constant-expression cannot be narrowed from type 'unsigned long long' to 'size_t' (aka 'unsigned int') in initializer list [-Wc++11-narrowing]
1    const size_t addrman_count{bench.epochs() * bench.epochIterations()};
2                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3bench/addrman.cpp:113:32: note: insert an explicit cast to silence this issue
4    const size_t addrman_count{bench.epochs() * bench.epochIterations()};
5                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
6                               static_cast<size_t>(                    )
71 error generated.
8make[2]: *** [Makefile:12508: bench/bench_bitcoin-addrman.o] Error 1

hebasto added the label Bug on Sep 4, 2021

hebasto commented at 1:32 pm on September 4, 2021: member

cc @jnewbery

MarcoFalke commented at 12:44 pm on September 8, 2021: member

@hebasto What are the steps to reproduce?

hebasto commented at 12:46 pm on September 8, 2021: member

Just to compile with all defaults on 32-bit hardware.

On Wed, 8 Sep 2021 at 15:44, MarcoFalke @.***> wrote:

@hebasto https://github.com/hebasto What are the steps to reproduce?

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub https://github.com/bitcoin/bitcoin/issues/22889#issuecomment-915206135, or unsubscribe https://github.com/notifications/unsubscribe-auth/AH3PXPSHIRNOON4Q2LQELA3UA5LDLANCNFSM5DNJRTHQ .

–

– Hennadii Stepanov

kristapsk commented at 12:58 pm on September 8, 2021: contributor

I think the Bitcoin software can safely drop support for 32-bit systems.

A lot of people run nodes on Raspberry Pi’s, which are nowadays 64-bit boards, but default OS (Raspbian) is still 32-bit.

MarcoFalke commented at 1:13 pm on September 8, 2021: member

I can reproduce on 64-bit via:

 0$ CONFIG_SITE="$PWD/depends/i686-pc-linux-gnu/share/config.site" ./configure CC="clang -m32" CXX="clang++ -m32"
 1$ make V=1
 2Making all in src
 3make[1]: Entering directory '/bitcoin-core/src'
 4make[2]: Entering directory '/bitcoin-core/src'
 5make[3]: Entering directory '/bitcoin-core'
 6make[3]: Leaving directory '/bitcoin-core'
 7/usr/bin/ccache clang++ -m32 -std=c++17 -DHAVE_CONFIG_H -I. -I../src/config  -fmacro-prefix-map=/bitcoin-core/src=. -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -I. -I./secp256k1/include  -I/bitcoin-core/depends/i686-pc-linux-gnu/include -I./leveldb/include -I./leveldb/helpers/memenv -I./univalue/include -I/bitcoin-core/depends/i686-pc-linux-gnu/include -pthread -I/bitcoin-core/depends/i686-pc-linux-gnu/include -I./bench/ -I/bitcoin-core/depends/i686-pc-linux-gnu/include/  -DHAVE_BUILD_INFO -D__STDC_FORMAT_MACROS -D_FILE_OFFSET_BITS=64 -DPROVIDE_FUZZ_MAIN_FUNCTION -fdebug-prefix-map=/bitcoin-core/src=. -Wstack-protector -fstack-protector-all -fcf-protection=full      -fPIE -pipe -O2  -MT bench/bench_bitcoin-addrman.o -MD -MP -MF bench/.deps/bench_bitcoin-addrman.Tpo -c -o bench/bench_bitcoin-addrman.o `test -f 'bench/addrman.cpp' || echo './'`bench/addrman.cpp
 8bench/addrman.cpp:113:32: error: non-constant-expression cannot be narrowed from type 'unsigned long long' to 'size_t' (aka 'unsigned int') in initializer list [-Wc++11-narrowing]
 9    const size_t addrman_count{bench.epochs() * bench.epochIterations()};
10                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
11bench/addrman.cpp:113:32: note: insert an explicit cast to silence this issue
12    const size_t addrman_count{bench.epochs() * bench.epochIterations()};
13                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
14                               static_cast<size_t>(                    )
151 error generated.

MarcoFalke commented at 1:21 pm on September 8, 2021: member

The following diff fixes it:

 0diff --git a/src/bench/addrman.cpp b/src/bench/addrman.cpp
 1index 8fbb68c04..17f54428e 100644
 2--- a/src/bench/addrman.cpp
 3+++ b/src/bench/addrman.cpp
 4@@ -6,6 +6,7 @@
 5 #include <bench/bench.h>
 6 #include <random.h>
 7 #include <util/time.h>
 8+#include <util/check.h>
 9 
10 #include <optional>
11 #include <vector>
12@@ -110,7 +111,8 @@ static void AddrManGood(benchmark::Bench& bench)
13      * we want to do the same amount of work in every loop iteration. */
14 
15     bench.epochs(5).epochIterations(1);
16-    const size_t addrman_count{bench.epochs() * bench.epochIterations()};
17+    const uint64_t addrman_count{bench.epochs() * bench.epochIterations()};
18+    Assert(addrman_count==5);
19 
20     std::vector<std::unique_ptr<CAddrMan>> addrmans(addrman_count);
21     for (size_t i{0}; i < addrman_count; ++i) {

hebasto commented at 1:22 pm on September 8, 2021: member

@bitcoin-public-domain

I wonder on what device you trying to compile the Bitcoin software.

ODROID-HC1

jonatack commented at 1:25 pm on September 8, 2021: member

Seems similar to #22464?

hebasto commented at 1:52 pm on September 8, 2021: member

Seems similar to #22464?

My concerns are about the “32-bit + dash [gui] [CentOS 8]” task on Cirrus CI which fails to catch both #22459 and this one.

MarcoFalke commented at 1:55 pm on September 8, 2021: member

That one isn’t using clang

hebasto commented at 1:57 pm on September 8, 2021: member

That one isn’t using clang

#22459 was caught with GCC 8.3.0.

MarcoFalke commented at 5:09 pm on September 8, 2021: member

#22923

laanwj closed this on Sep 9, 2021

DrahtBot locked this on Oct 30, 2022

Compiling fails on 32-bit systems #22889