Mild NACK I prefer noisy failures. If the failures here are burdensome, can you cite some?

@Empact How is this less noisy than before? :-)

-    format(oss, fmt.c_str(), args...);
+    try {
+        tfm::format(oss, fmt.c_str(), args...);
+    } catch (const tinyformat::format_error& fmterr) {
+        fprintf(stderr, "Format error: %s (fmt=\"%s\")\n", fmterr.what(), fmt.c_str());
+        std::terminate();
+    }

Basically, blow up when possible, return a value or log or both when not possible. Make errors as difficult to ignore as possible.

@Empact I agree, but I don't get how this is less noisy. std::terminate() is called earlier :-)

Fair enough, terminate is effectively noisy, but given terminate would be called if the exception were uncaught, ironically, this now makes errors which would have previously been blanket-caught via catch(...) etc a terminate which is a potentially problematic behavior change. But maybe I'm wrong on that? It was called out on me here relative to converting asserts to throw: #13268

FTR I don't like that we have blanket-catches in the codebase, but they're sometimes warranted.

Hopefully we will have format string linting soon (see separate PR) and perhaps that will allow us to make strprintf(…) noexcept. But I'm not sure std::terminate is the right way to go here in case of formatting errors. Perhaps better to just print an error message to stderr and/or return the error message to the caller? @laanwj, what do you think?

• Move src/tinyformat.h to src/tinyformat/tinyformat.h #13846 (Move src/tinyformat.h to src/tinyformat/tinyformat.h by Empact)
• Include tinyformat as a subtree #13845 (Include tinyformat as a subtree by Empact)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

I'm going to close this, I don't think this change is really necessary.