On Mon, 30 Mar 2026 12:36:55 GMT, Thomas Stuefe <[email protected]> wrote:

> Some small QOL improvement and trivial fixes:
> - System.dump_map now takes the filename as a positional argument, which 
> makes for less awkward handling
> - System.dump_map now allows %t as a placeholder in its filename, and the 
> default filename now contains a timestamp (reusing the logic in 
> `make_log_name` for this)
> - `make_log_name` returns a C-heap allocated array and therefore should not 
> return a `const char*` but a `char*`.
> - Used the more common - in our code base - `_WINDOWS` instead of `_WIN64`
> - Removed the "(linux, Windows and MacOS only)" output since this code never 
> runs on other platforms (not compiled for AIX or *BSD)
> - since the filename argument has a default value if omitted by the user, it 
> can never be empty - assert that instead of doing runtime checks

I hope I didn't slow this down.  This is of course good.
System.dump_map was the only command that took "-F filename" to specify an 
output file, or even used an option for an output filename I think, so good to 
remove that (and maybe in future we can avoid multiple commands just because 
they may write to a file).

Would be great to update src/jdk.jcmd/share/man/jcmd.md , including to add %t 
in the filename description for this jcmd.

The default filename doesn't have to include %t but I think you're suggesting 
it's required often enough to be the default (people frequently make multiple 
dump_map calls over time).

This does change a command-line interface to a jdk/bin command so a CSR should 
really be used.  It should be routine and uncontroversial, it's making the 
interface more consistent.  I think that would be for the removal of -F only.

-------------

PR Review: https://git.openjdk.org/jdk/pull/30497#pullrequestreview-4083540789

Reply via email to