it->dir

it->filterExpression/filter/expression?

Why change away from auto?

it -> mountPointPtr? Perfect would be renaming the existing mountpoint to mountPointPath or something, so mountPoint` could be used as loop var name. But I also think this would be too invasive WRT commit history and commit purpose.
Or mp, as used in some other code also looping over the same structure, see below in src/widgets/kpropertiesdialog.cpp

it->mountPoint

it->desktopFilePath (hml, that recentDocuments() is semantically quite misleading without reading the docs).

Curious: while touching this, would a static const or constexpr not be the even better option?
Any rule of thumb known when telling the compiler which way to treat this?
Right now unsure what the compiler even does with a plain const array in a method without any optimization. It copies the array data onto the stack on the method invocation, or?

missing space after for

it->info

it -> device

unrelated change, please separate commit

Unrelated change/fix, please separate commit. Though, is this even correct? There is no QString::operator>=(QChar) & similar, is there?

Unrelated, please separate commit. Was the code checked that isNull() is not relevant?

While touching all the lines, would improve the code to cache match.value() into some
QString& matchString = match.value(); IMHO.

I would make this a separate commit, as this does not exactly match the commit message and might confuse future history readers.

while touching all the lines & vars, desktopEntryName() could be cached.

I would make this a separate commit, as this does not exactly match the commit message and might confuse future history readers.

I would make this a separate commit, as this does not exactly match the commit message and might confuse future history readers.

I would make this a separate commit, as this does not exactly match the commit message and might confuse future history readers.

I would make this a separate commit, as this does not exactly match the commit message and might confuse future history readers.

Consistency with all other range for loops in kio, and https://wiki.qt.io/Coding_Conventions#auto_Keyword

picked mp

Good question.

Without optimizations, I assume static means generating atomics for thread-safety, while on stack generates a memcpy indeed.
I checked the assembly on godbolt.org, for that last part. Couldn't see atomics with static because the testcase is all in main(), so no threads involved, but in a library method this is a requirement.

With -O2, the generated code is exactly the same.
https://godbolt.org/z/Ez8fba

So I guess it doesn't matter?

Indeed, this whole subdir was in need of code reformatting.
Done, https://commits.kde.org/kio/c8c2eff2a2362278acddcc995221176a7788e24e

Oh indeed, I thought c was a QChar, given its name...

I guess this creates a QString from the QChar but yeah I'll revert.

the code here uses clear() already for ql elements.

and this is the final use of ql, null vs empty makes no difference =>
https://commits.kde.org/kio/ea035cd204dea8cd563155c4256400fe67e8b69c

Moved all overrides to https://commits.kde.org/kio/54429f8a36aad65088b02eb9d13c3bf78f61f22f

https://commits.kde.org/kio/1660e30f3821204b08b43910738fa52ce1cf0915