[sahvx655-wq]

Description

Carrying on from #5265 I went back over the t-string paths looking for other f-string
guards that had not been extended, and is_docstring is one. It rejects bytes and
f-strings with set(prefix).intersection("bBfF"), a list written well before PEP 750, so
a t-string sitting as the first statement of a module, class or function is classified as
a docstring and visit_STRING runs the whole docstring rewrite over it: .strip() for a
single-line literal, fix_multiline_docstring for a triple-quoted one. What that touches
is not indentation inside a docstring but the content of a runtime value, so
t" spam " comes back as t"spam" and the Template.strings it builds goes from
(' spam ',) to ('spam',). Python never takes a t-string as __doc__
(ast.get_docstring returns None, and so does f.__doc__), so there was never
anything here to reindent.

On 3.14 the AST safety check does notice, so under the default --safe the symptom is
INTERNAL ERROR: ... produced code that is not equivalent to the source. Please report a bug and the file cannot be formatted at all, while --fast, blackd with
X-Fast-Or-Safe: fast and a direct format_str skip that check and write the shortened
template out silently, which is the half I find harder to live with. The test belongs in
is_docstring rather than in visit_STRING because the blank-line handling in
lines.py reaches the same predicate through Line.is_docstring, and pushing it to the
call sites would mean repeating the prefix check in each of them. An f-string against
t-string differential over 18 constructs in stable and preview is how I satisfied myself
this was the last gap: 20 divergences before the patch, none after. Formatting black's
own src, tests, scripts and docs in stable, preview, --skip-string-normalization
and pyi modes gives 1356 results and the only one that moves is the new case. One thing
to flag while I am here: T stays uppercase in the expected output because
normalize_string_prefix lowercases F and B but not T, which reads like a separate
oversight and I have kept it out of this patch.

Checklist - did you ...

• Implement any code style changes under the --preview style, following the stability policy?
• Add an entry in CHANGES.md if necessary?
• Add / update tests if necessary?
• Add new / update outdated documentation?

[github-actions]

diff-shades results comparing this PR (df78275) to main (c7801d9):

--preview style: no changes

--stable style: no changes

---

What is this? | Workflow run | diff-shades documentation

[JelleZijlstra]

Looks good, thanks.

Is your reasoning for applying this directly to the stable style that this code would have crashed otherwise, so it doesn't make sense to fix it only in preview?

[tusharsadhwani]

@JelleZijlstra yeah, currently only --fast lets this pass through silently