Currently KProcessList had a method called "name" which confusingly
returns the entire command line.
In many cases where we want to use this class we need both. (See D22327)
This class exposes both "name" and "command" with the respective
properties.
In order to do this correctly we need to split the process name before
the null characters are replaced with spaces.
Unit test
Manually invoked ps list output
| Automatic diff as part of commit; lint not applicable. |
| Automatic diff as part of commit; unit tests not applicable. |
+1
| src/lib/util/kprocesslist_p.h | ||
|---|---|---|
| 46 | Order name before since that's how the methods are ordered | |
So now name() will often return an empty string?
| src/lib/util/kprocesslist_unix.cpp | ||
|---|---|---|
| 142 | if it's -1, when you do ++ it will become 0 anyway. | |
So now name() will often return an empty string?
That's certainly not the intention.
The system is:
Firstly we do the fallback. Reading from /proc/N/state . Though this is rubbish and truncates things.
Both name and command are set to this.
Then we read /proc/N/cmdline
Then we extract the command.
We set the name to the first segment of the command (between the final / and the first null byte
We set the command to the full thing, but with all null bytes becoming spaces
On windows, name and command will return the same thing.
| src/lib/util/kprocesslist_unix.cpp | ||
|---|---|---|
| 142 | It would, but it would then requires a huge comment explaining it. | |
Should we ship this now? After frameworks tagging I guess.
Also, can you add a CHANGELOG: KProcessInfo::name() now returns only the name of the executable. For the full command line use KProcessInfo::command() so it shows up nicely in the release announcement changelog