32.0 RC Testing Guide Feedback #36258

janb84 commented at 8:36 AM on September 15, 2026: contributor

This issue is to discuss the 32.0 Release Candidate Testing Guide. If you have any feedback on the document, please leave a comment here.

Note: This is for feedback on the document, not on Bitcoin Core or on the 32.0 changes. Create a new issue or give feedback in the global issue here.

Thank you for taking a look at the guide and leaving your feedback.

Sjors commented at 9:41 AM on September 15, 2026: member

Guix builders: note that for v32.0rc1 the guix-codesign step will fail. See #36256. But you still test everything else.

hebasto commented at 9:46 AM on September 15, 2026: member

Guix builders: note that building on RISC-V will fail. See #36233.

fanquake added this to the milestone 32.0 on Sep 15, 2026

epiccurious commented at 4:30 PM on September 15, 2026: contributor

On the 32.0 RC Testing Guide doc, section 5.4.1's title should be updated to V32 Aliases. Also update the alias from alias bcli31=... to alias bcli32=....

https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#541-v31-aliases

Also, in Section 4 there's a reference to a previous version v31.2 that should be corrected to v31.1.

https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#4-obtain-previous-release

tankyleo commented at 11:54 PM on September 16, 2026: none

Successfully rebuilt all indices, sizes measured using du -h -d 1. Before:

91G	./txospenderindex
69G	./txindex
13G	./blockfilter
172G	.

After:

83G	./txospenderindex
23G	./txindex
13G	./blockfilter
118G	.

Sjors commented at 11:36 AM on September 21, 2026: member

I ran into another issue with guix-codesign for rc2, this time App Armor related:

zcat /var/log/guix/drvs/2s/v5049dk6ifphfm20xp7r0rdc1vm2bv-python-signapple-3fab3bb57f227f0dd31007b417683035f5204838-checkout.drv.gz
warning: unable to access '/etc/gitconfig': Permission denied
warning: unable to access '/etc/gitconfig': Permission denied
warning: unable to access '/etc/gitconfig': Permission denied
fatal: unknown error occurred while reading the configuration files
git-fetch: '/gnu/store/xz7xygq040vx78snmla70y03h6x9yypg-git-minimal-2.52.0/bin/git init --initial-branch=main' failed with exit code 128

According to Claude, and my own testing, adding this inside the profile guix-helper { … } block in /etc/apparmor.d/guix-daemon, helps:

# builtin:git-download runs git unchrooted; git treats EACCES on the
# system config as fatal (ENOENT would be fine).
/etc/gitconfig r,

I haven't filed an upstream issue yet.

yuvicc commented at 8:32 AM on September 30, 2026: contributor

In https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#11-exportwatchonlywallet-new-rpc, the last part:

Stop the old node and clean up:

bcli32 stop && sleep 5  &&  datadir-cleanup

should be bcli32 instead of bcli31

janb84 commented at 10:32 AM on September 30, 2026: contributor

Thank you for the feedback

In https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#11-exportwatchonlywallet-new-rpc, the last part:

Stop the old node and clean up:

bcli32 stop && sleep 5 && datadir-cleanup should be bcli32 instead of bcli31

Mhh no, we are stopping the old node, that is intentional. a few lines above we stop the v32 node and start importing the watchonlywallet on the "old" V31 node. Have change the wording a bit to clarify.

Similarly for https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#12-exportasmap-new-rpc and https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#13-getopenrpcinfo-and-rpcdiscover should be bcli32.

This one should indeed be V32, have corrected it. TY.

brunoerg commented at 1:15 PM on September 30, 2026: contributor

The expected output for exportasmap in the guide is:

{
  "path": "/path/to/DATA_DIR_32/asmap.dat",
  "bytes_written": 1582208,
  "file_hash": "..."
}

I think instead of "..." we could put the real file hash since everybody - following the guide - should get the same hash (03580ade8ec0036ad3d6a5a91602b995c89387f722d3fee58127219de6aafc12).

Also, "Check that the file exists and isn't empty" runs ls -l so we could tell people to compare the bytes from this command to the one returned by the RPC.

erickcestari commented at 1:42 PM on September 30, 2026: none

Shouldn't we export the ASMAP from bitcoind32 and import it into bitcoind31 to verify the ASMAP import and check whether the ASMAP version hash in the log matches the file_hash?

I tried this, and the ASMAP version log shows a different hash when importing the same ASMAP into bitcoind32 and bitcoind31.

[erick@archlinux rctesting]$ bcli32 exportasmap $DATA_DIR_32/asmap.dat
{
  "path": "/tmp/32-rc-test/asmap.dat",
  "bytes_written": 1582208,
  "file_hash": "03580ade8ec0036ad3d6a5a91602b995c89387f722d3fee58127219de6aafc12"
}

ViniciusCestarii commented at 1:59 PM on September 30, 2026: contributor

At 2.2 PSBT v2 default it says:

The output should be a transaction ID (txid): 89be67d3cdf04192c718e6f1e6486b02172c53f5f39b0ab73de1c53b0707f1e3

This could mislead readers into thinking their txid should exactly match this value. It would be better to indicate that this is an example txid, rather than the expected output.

brunoerg commented at 2:23 PM on September 30, 2026: contributor

The expected output in the guide for the estimatesmartfee for each mode is (https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#25-estimatesmartfee-combined-estimator):

none=0.00000100 block_policy=0.00004000 mempool_policy=0.00000100

but shouldn't it be:

none=0.00000100 block_policy=0.00003000 mempool_policy=0.00000100

sedited commented at 2:32 PM on September 30, 2026: contributor

I tried this, and the ASMAP version log shows a different hash when importing the same ASMAP into bitcoind32 and bitcoind31.

That is intentional. The hash calculation in v32 was changed to be compatible with sha256sum on the command line.

janb84 commented at 2:48 PM on September 30, 2026: contributor

he expected output for exportasmap in the guide is:

{ "path": "/path/to/DATA_DIR_32/asmap.dat", "bytes_written": 1582208, "file_hash": "..." } I think instead of "..." we could put the real file hash since everybody - following the guide - should get the same hash (03580ade8ec0036ad3d6a5a91602b995c89387f722d3fee58127219de6aafc12).

Also, "Check that the file exists and isn't empty" runs ls -l so we could tell people to compare the bytes from this command to the one returned by the RPC.

Corrected.

The expected output in the guide for the estimatesmartfee for each mode is (https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#25-estimatesmartfee-combined-estimator):
none=0.00000100 block_policy=0.00004000 mempool_policy=0.00000100
but shouldn't it be:

none=0.00000100 block_policy=0.00003000 mempool_policy=0.00000100

Output is what I got after the commands, will run it again soon (tm). If output is the same I will add a note that the output can be slightly different.

janb84 commented at 2:53 PM on September 30, 2026: contributor

At 2.2 PSBT v2 default it says:

The output should be a transaction ID (txid): 89be67d3cdf04192c718e6f1e6486b02172c53f5f39b0ab73de1c53b0707f1e3

This could mislead readers into thinking their txid should exactly match this value. It would be better to indicate that this is an example txid, rather than the expected output.

corrected.

janb84 commented at 2:59 PM on September 30, 2026: contributor

That is intentional. The hash calculation in v32 was changed to be compatible with sha256sum on the command line.

I have added a note to the section with this explanation. Thanks for clarifying @sedited

jcastros commented at 1:19 PM on October 3, 2026: contributor

A couple of mismatches on the published v32.0rc2 x86_64 binary.

In 1.3, both of these return 157, not 156:

bcli32 getopenrpcinfo | jq '.methods | length'
bcli32 help | grep -c '^[a-z]'

With show_hidden=true the count is 176, not 175.

In 2.5, the guide lists fee_reason as "fee rate estimator", "mempool minimum", "fallback", or "minimum required". Those are not the strings. The possible values, from StringForFeeReason, are "Fee Rate Estimator", "Mempool Min Fee", "User Specified Fee", "Fallback fee", and "Minimum Required Fee".

sendtoaddress with fee_rate=5 and verbose=true returned "User Specified Fee". After the fee-history steps in that section, a send with no fee_rate returned "Minimum Required Fee". That is the output the guide should show for that command. I didn't hit the other three values.

jaonoctus commented at 9:03 AM on October 4, 2026: none

Small notes on the guide:

2.5: ~+1 to @brunoerg~ , I got block_policy=0.00003000. Since fee rates are random, maybe mark that output as an example.
2.5: (nit) fee_reason returns "Minimum Required Fee", while the guide lists lowercase names.
Step 6: when building from a git checkout, --version shows v32.0rc2 (git tag) while getopenrpcinfo shows v32.0.0rc2. That's expected but could confuse source builders.

janb84 commented at 12:57 PM on October 4, 2026: contributor

In 1.3, both of these return 157, not 156:

Corrected, not sure why the off by one error was in there.

"Fee Rate Estimator", "Mempool Min Fee", "User Specified Fee", "Fallback fee", and "Minimum Required Fee".

Have corrected this, was direct copy from the release notes. I also corrected this in the release notes.

Step 6: when building from a git checkout, --version shows v32.0rc2 (git tag) while getopenrpcinfo shows v32.0.0rc2. That's expected but could confuse source builders.

Have added additional note.

The expected output in the guide for the estimatesmartfee for each mode is (https://github.com/bitcoin-core/bitcoin-devwiki/wiki/32.0-Release-Candidate-Testing-Guide#25-estimatesmartfee-combined-estimator):
none=0.00000100 block_policy=0.00004000 mempool_policy=0.00000100
but shouldn't it be:

none=0.00000100 block_policy=0.00003000 mempool_policy=0.00000100

So I have re-run this on my machine aarch64/nix and x86-64/debian 13 but both return (consistently) the numbers in the guide, I used the dev-mode preset. Can you report the build setting ?

jaonoctus commented at 1:57 PM on October 5, 2026: none

So I have re-run this on my machine aarch64/nix and x86-64/debian 13 but both return (consistently) the numbers in the guide, I used the dev-mode preset. Can you report the build setting ?

My guess it's from fee_rate=$((RANDOM % 50 + 1)), it makes the block policy estimate vary from run to run.

build setting:

Source: v32.0rc2 (bc795e60db)
Configure: plain cmake -B build
Features on: wallet, IPC, external signer, embedded ASMap, tests
Features off: GUI, ZMQ, USDT, bench, fuzz
Compiler: Apple clang 21.0.0 (clang-2100.3.34.2)
SDK: macOS 27.0
OS: macOS 26.6.2, arm64

janb84 commented at 11:26 AM on October 11, 2026: contributor

So I have re-run this on my machine aarch64/nix and x86-64/debian 13 but both return (consistently) the numbers in the guide, I used the dev-mode preset. Can you report the build setting ?

My guess it's from fee_rate=$((RANDOM % 50 + 1)), it makes the block policy estimate vary from run to run.

build setting:

Source: v32.0rc2 (bc795e6)

Configure: plain cmake -B build

Features on: wallet, IPC, external signer, embedded ASMap, tests

Features off: GUI, ZMQ, USDT, bench, fuzz

Compiler: Apple clang 21.0.0 (clang-2100.3.34.2)

SDK: macOS 27.0

OS: macOS 26.6.2, arm64

Ah yes that makes sense, forgot about that part. Have added some additional notes. Thanks

fanquake commented at 8:03 AM on October 14, 2026: member

A note for future testing guides, ideally the guide would have caught/surfaced this earlier: #36518.