Port endl as in qt5.15 it's namespaced
Details
Details
Diff Detail
Diff Detail
- Repository
- R237 KConfig
- Branch
- port_endl (branched from master)
- Lint
No Linters Available - Unit
No Unit Test Coverage - Build Status
Buildable 20640 Build 20658: arc lint + arc unit
| autotests/kconfigtest.cpp | ||
|---|---|---|
| 533–542 | When it has only strings you can use R"()" | |
Comment Actions
| autotests/kconfigtest.cpp | ||
|---|---|---|
| 533–542 | What would be the benefit here? One double-backslash simplified to a single one, and that's it, right? | |
| src/kconfig_compiler/kconfig_compiler.cpp | ||
| 694 | Why? Surely the QTextStream destructor (which is called here) will take care of that. | |
| 729 | (same) | |
| 881 | (pre-existing) Strange that this one is cout and not cerr... | |
| 1379 | I've been wondering if this flush() is needed. In my tests, it works without. http://www.davidfaure.fr/2020/qtextstream_flush.cpp But in theory local variables are deleted *after* the return value is copied (otherwise mutex lockers wouldn't make sense, stack overflow confirms) so in the absence of named-return-value-optimization (which IIRC is only guaranteed in C++17 and later), the copying could happen before the flushing. Unless I missed something, let's keep the flush() to be safe :-) | |
| src/kconfig_compiler/kconfig_compiler.cpp | ||
|---|---|---|
| 694 | I think that it was a problem when QTextStream used QString which doesn't flush it. | |
| src/kconfig_compiler/kconfig_compiler.cpp | ||
|---|---|---|
| 694 | I think you're confusing two cases. The "underlying device" goes away before the QTextStream ==> this only works when the device is a proper QIODevice, while a QString has no "closed notification". The QTextStream goes away before the underlying device. No problem in either case, the QTextStream destructor flushes. | |