init: log resolved datadir path #36373

LarryRuane commented at 5:42 PM on September 28, 2026: contributor

Because the data directory may be a symlink, also log the path it resolves to, so debug.log files show the actual location for that run.

This patch would mostly help developers. I've been testing a potential performance PR across different data directories (for example, SSD versus HDD). After each run, I save the debug.log file. It would be nice if the debug.log file showed which data directory was used for that run.

There are two ways to change which data directory to use: -datadir= on the command line, or by making the default path a symlink. If a script starts bitcoind (e.g., on an RPi with a whole infrastructure, like Umbrel or myNode), changing the startup command arguments may be inconvenient or not even possible. It's easy to change the default directory to be a symlink. But in that case, debug.log doesn't show where the actual data directory is located. This patch adds one line to the logging startup output showing the resolved data directory path (after following symlinks).

I don't think a separate test is needed. I haven't tested on all platforms, but this patch uses primitives already used elsewhere, so it should work everywhere.

Here's an example of the additional line this patch logs during startup, just after the Using data directory line (omitting the timestamp prefix). I'll show both here for context (only the second line is added):

Using data directory /home/larry/.bitcoin
Resolved data directory /sd/.bitcoin

The patch could print this line only if the resolved path differs from the Using data directory path, but we may as well print it every time, because it's only one extra line, and this way any scripts that look for the resolved path are simpler (they can depend on it always being there).

DrahtBot commented at 5:42 PM on September 28, 2026: contributor

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

External sites

Code coverage & benchmarks: https://corecheck.dev/bitcoin/bitcoin/pulls/36373
Ralph review <img src=https://review.fish.foo/assets/bitcoin/bitcoin/36373.png alt="Ralph review status"> https://git.fish.foo/bitcoin/bitcoin/pulls/36373#author-ralph

Reviews

See the guideline and AI policy for information on the review process. A summary of reviews will appear here.

init: log resolved data directory path

Log a separate line with the data directory after resolving symlinks,
so saved debug.log files show where the data actually was for that run.
The existing "Using data directory" line is left unchanged so scripts
that parse it aren't affected. The new line is always printed, even when
the path doesn't change, so scripts can rely on it being there.

21e37651cc

LarryRuane force-pushed on Oct 7, 2026

in src/init/common.cpp:132 in 21e37651cc

 124 | @@ -124,7 +125,11 @@ bool StartLogging(const ArgsManager& args)
 125 |          LogInfo("Startup time: %s", FormatISO8601DateTime(GetTime()));
 126 |      }
 127 |      LogInfo("Default data directory %s", fs::PathToString(GetDefaultDataDir()));
 128 | -    LogInfo("Using data directory %s", fs::PathToString(gArgs.GetDataDirNet()));
 129 | +    const fs::path datadir{gArgs.GetDataDirNet()};
 130 | +    LogInfo("Using data directory %s", fs::PathToString(datadir));
 131 | +    std::error_code ec;
 132 | +    const fs::path real_datadir{fs::canonical(datadir, ec)};
 133 | +    LogInfo("Resolved data directory %s", fs::PathToString(ec ? datadir : real_datadir));

l0rinc commented at 1:19 AM on October 8, 2026:

The commit message says the resolved path is "always printed", but the PR title says "when it differs".

Is this part of a broader effort to improve symlink support? If so, I wouldn't want to see two separate log lines for the exact same folder in 99% of cases.

LarryRuane commented at 2:13 PM on October 8, 2026:

Thanks for taking a look. The first version of this PR printed the resolved path at the end of the same line that exists today:

- Using data directory /home/larry/.bitcoin
+ Using data directory /home/larry/.bitcoin (/sd/.bitcoin)

And it added that path only if it was different from the "Using" path. But after more thought, it seems better to log a new line ("Resolved") and leave the existing log line unchanged (in case scripts exist somewhere that expect the current format). I also thought it would be better to always log this new Resolved line so future scripts can look for it without needing logic to fall back to "Using" if "Resolved" doesn't exist. So I force-pushed those changes. But I forgot to update the title; fixed now.

Sounds like you would like the Resolved line to only print if it's different; I can make that change, I don't feel strongly either way.

No, I'm not thinking this is part of a broader effort to improve symlink support. AFAIK symlinks are already fully supported. It's just a self-contained PR that would have helped me personally, since I saved a bunch of debug.log files and later wasn't sure whether each corresponded to an SSD or an HDD data directory run (because I had changed a symlink). Plus, it seems generally useful to know where the actual data directory is.

I can make the logging of that Resolved line conditional so it doesn't print the same path as the Using line, maybe I'll wait for a few more opinions.

LarryRuane renamed this:
~~init: log resolved datadir path when it differs~~
init: log resolved datadir path
on Oct 8, 2026