This was added to work around boost::filesystem throwing when the environments locale was invalid. Boost.Filesystem has been removed, and std::filesystem doesn't use the locale on POSIX (and bitcoind never calls setlocale).
Qt calls setlocale(LC_ALL, "") on startup, and falls back to C.UTF-8 if the locale isn't UTF-8.
Alternative to #36399, which extends the hack further.
<!--e57a25ab6845829454e8d69fc972939a-->
The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.
<!--006a51241073e994b41acfe9ec718e94-->
<!--021abf342d371248e50ceaed478a90ca-->
See the guideline and AI policy for information on the review process.
| Type | Reviewers |
|---|---|
| ACK | maflcko, stickies-v, sedited |
If your review is incorrectly listed, please copy-paste <code><!--meta-tag:bot-skip--></code> into the comment that the bot should ignore.
<!--5faf32d7da4f0f540f40219e4f7537a3-->
Qt calls
setlocale(LC_ALL, "")on startup, and falls back toC.UTF-8if the locale isn'tUTF-8.
Can you add a reference/source to this claim?
Can you add a reference/source to this claim?
We also claim this in: https://github.com/bitcoin/bitcoin/blob/60ebd98c79043e99c582bcbaf412bc424c8daf9b/src/qt/bitcoin.cpp#L203
lgtm ACK 622df9e941dd55fc9b01125ea1de659ed6a9794e
ACK 622df9e941dd55fc9b01125ea1de659ed6a9794e
Doesn’t this change affect the behaviour of the scripts specified via the -*notify options in environments where the locale is invalid?
Yes, that is also what Ralph says (https://git.fish.foo/bitcoin/bitcoin/pulls/36524#author-ralph) , but I think users are better off not setting an invalid locale. If they really want to set an invalid locale, they are better off setting the locale in their notify script explicitly. (Explicitly setting it works before and after the changes here)
69 | - // may be invalid, in which case the "C.UTF-8" locale is used as fallback. 70 | -#if !defined(WIN32) && !defined(__APPLE__) && !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__) 71 | - try { 72 | - std::locale(""); // Raises a runtime error if current locale is invalid 73 | - } catch (const std::runtime_error&) { 74 | - setenv("LC_ALL", "C.UTF-8", 1);
This surfaces a bug in interface_gui.py, that passes on master but fails on 622df9e941dd55fc9b01125ea1de659ed6a9794e.
On CI, we set LC_ALL=C.UTF-8 in all scripts, so the locale is set in all tests, which means that we're not really testing this change.
When running this test locally on a Linux environment with LC_ALL unset and no nl_NL.UTF-8 locale installed, the interface_gui.py test fails. This is because in test_node.py:302, we set subp_env.setdefault("LC_ALL", "nl_NL.UTF-8"). On CI, this operation is a no-op because LC_ALL is always set, so the nl_NL.UTF-8 locale is never actually set. When it is actually set, and nl_NL.UTF-8 isn't installed on the machine, the test fails.
Tested in a debian:trixie container:
$ python3 build/test/functional/interface_gui.py
...
2026-10-14T14:57:03.224515Z TestFramework (INFO): PRNG seed is: 6521282163018713296
2026-10-14T14:57:03.276866Z TestFramework (INFO): Initializing test directory /tmp/bitcoin_func_test_nr9vej0d
2026-10-14T14:57:03.871440Z TestFramework (INFO): Test that bitcoin-gui starts up and can be stopped via RPC
2026-10-14T14:57:04.243870Z TestFramework (ERROR): Unexpected exception:
Traceback (most recent call last):
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/test_framework.py", line 146, in main
self.run_test()
~~~~~~~~~~~~~^^
File "/Users/stickies-v/code/bitcoin/review/36524/build/test/functional/interface_gui.py", line 34, in run_test
self.stop_node(0)
~~~~~~~~~~~~~~^^^
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/test_framework.py", line 527, in stop_node
self.nodes[i].stop_node(expected_stderr, wait=wait)
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/test_node.py", line 517, in stop_node
self.wait_until_stopped(expected_stderr=expected_stderr)
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/test_node.py", line 563, in wait_until_stopped
self.wait_until(lambda: self.is_node_stopped(**kwargs), timeout=timeout)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/test_node.py", line 913, in wait_until
return wait_until_helper_internal(test_function, timeout=timeout, timeout_factor=self.timeout_factor, check_interval=check_interval)
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/util.py", line 449, in wait_until_helper_internal
if predicate():
~~~~~~~~~^
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/test_node.py", line 563, in <lambda>
self.wait_until(lambda: self.is_node_stopped(**kwargs), timeout=timeout)
~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^
File "/Users/stickies-v/code/bitcoin/review/36524/test/functional/test_framework/test_node.py", line 548, in is_node_stopped
raise AssertionError("Unexpected stderr {} != {}".format(stderr, expected_stderr))
AssertionError: Unexpected stderr Detected locale "C" with character encoding "ANSI_X3.4-1968", which is not UTF-8.
Qt depends on a UTF-8 locale, and has switched to "C.UTF-8" instead.
If this causes problems, reconfigure your locale. See the locale(1) manual
for more information.
The test runs fine with LC_ALL=C.UTF-8 python3 build/test/functional/interface_gui.py.
From #35551, it seems like we set a well-translated language to ensure that in a foreign locale, RPC strings still come through in English. If that's what we want to achieve, then perhaps we could e.g. add https://github.com/stickies-v/bitcoin/commit/a39fa5420ed965f8975221ce49bd813fac3aed61 as a minimal fix prior to this change?
I'm not completely sure what the expectations are with this test, but I've cherry-picked your commit here.
From #35551 (review):
Could also set something like
LC_ALL=nl_NL.UTF-8to check the RPC response is not translated by accident?
Concept ACK
This surfaces an already buggy test, that now fails on some local environments, see comment.
Only change is a gui test fixup.
review ACK 27be7413e82072e8e68b958e808bbee379f4ac15 🚇
<details><summary>Show signature</summary>
Signature:
untrusted comment: signature from minisign secret key on empty file; verify via: minisign -Vm "${path_to_any_empty_file}" -P RWTRmVTMeKV5noAMqVlsMugDDCyyTSbA3Re5AkUrhvLVln0tSaFWglOw -x "${path_to_this_whole_four_line_signature_blob}"
RUTRmVTMeKV5npGrKx1nqXCw5zeVHdtdYURB/KlyA/LMFgpNCs+SkW9a8N95d+U4AP1RJMi+krxU1A3Yux4bpwZNLvVBKy0wLgM=
trusted comment: review ACK 27be7413e82072e8e68b958e808bbee379f4ac15 🚇
zmTJxrqt5CAm6rIQ+TMaGY6wW+CucsSb5e19C8wSwr0yqL18TX53qz6OFVwumCGNyfHrBbGmMx16i1yiR0xADw==
</details>
I mean if people feel strongly, there could be a release note along the lines:
The UTF-8 fallback for invalid or unavailable environment locales has been removed. Affected -*notify scripts may explicitly set a valid locale, such as LC_ALL=C.UTF-8, if needed.
ACK 27be7413e82072e8e68b958e808bbee379f4ac15
I mean if people feel strongly, there could be a release note along the lines:
I actually think that would be good. Shouldn't affect many people, but probably some, and having a mention in the release notes may save them a lot of confusion. Your proposed text works well for me.
This changes test behavior: GUI nodes now run with the Dutch
translation on all systems, also in CI. Before, they ran in English in
CI, because CI exports LC_ALL=C.UTF-8 and setdefault("LC_ALL",
"nl_NL.UTF-8") did nothing. So the translation check never ran there.
The old setting also fails when LC_ALL is not set and nl_NL.UTF-8 is
not installed: Qt prints a warning to stderr, and interface_gui.py fails
when SetupEnvironment() no longer replaces LC_ALL. -lang does not need
the locale to be installed. Date and number formats do not change.
This was added to work around `boost::filesystem` throwing when the
environments locale was invalid. Boost.Filesystem has been removed, and
`std::filesystem` doesn't use the locale on POSIX (and bitcoind never
calls `setlocale`).
Qt calls `setlocale(LC_ALL, "")` on startup, and falls back to C.UTF-8
if the locale isn't UTF-8.
Added a release note.
review ACK 43bb5e3c467f38566f7d7271b0030ef4af1cf78c 🚂
<details><summary>Show signature</summary>
Signature:
untrusted comment: signature from minisign secret key on empty file; verify via: minisign -Vm "${path_to_any_empty_file}" -P RWTRmVTMeKV5noAMqVlsMugDDCyyTSbA3Re5AkUrhvLVln0tSaFWglOw -x "${path_to_this_whole_four_line_signature_blob}"
RUTRmVTMeKV5npGrKx1nqXCw5zeVHdtdYURB/KlyA/LMFgpNCs+SkW9a8N95d+U4AP1RJMi+krxU1A3Yux4bpwZNLvVBKy0wLgM=
trusted comment: review ACK 43bb5e3c467f38566f7d7271b0030ef4af1cf78c 🚂
0RN8toZudX3WfXDZvsUbFkOd9P5Kd1le5Ef2zAGNlR9jHQMC0uLLAW10sHCvqSo5XvCP6+k/ZD/wjaINsvKADA==
</details>
ACK 43bb5e3c467f38566f7d7271b0030ef4af1cf78c
ACK 43bb5e3c467f38566f7d7271b0030ef4af1cf78c