On Tue, 2 Mar 2021 05:18:51 GMT, Chris Plummer <cjplum...@openjdk.org> wrote:
>>> You'll need a CSR for this. >> >> Ok, I will add it after this discussion ( `coonect` or `attach` ) >> >>> Will `jhsdb clhsdb --connect debugserver` work? If not, it should to be >>> consistent with the other commands. >> >> It does not work. CLHSDB and HSDB do not accept '--connect'. I plan to add >> it after this PR. >> >>> Does `detach` undo the `connect`? >> >> Yes. >> >>> It's not very symetric if we do since we already had `attach` + `detach` >>> and now you added `connect` + `detach`. Since you can `attach` to a pid or >>> a core file, maybe instead of adding `connect` you should just put the >>> debugd support in `attach`. >> >> I thought `connect` command is easy to understand because it is similar with >> `--connect` option in jhsdb, and also the change is smaller than adding to >> `attach` because `attach` already has 1 string argument (PID). >> I'm not familiar of network, but I concern hostname which is configured by >> numeric chars only - I'm not sure it is allowed, but [RFC >> 1123](https://tools.ietf.org/html/rfc1123) allows digit character at the >> first. >> >> The syntax of a legal Internet host name was specified in RFC-952 >> [DNS:4]. One aspect of host name syntax is hereby changed: the >> restriction on the first character is relaxed to allow either a >> letter or a digit. Host software MUST support this more liberal >> syntax. >> >> If digit hostname (e.g. 1234) is allowed, we cannot add debugd support to >> `attach` because we cannot distinguish it is PID or hostname. If so, it is >> reasonable to add `connect` command for this purpose. > > `jhsdb` has `--pid`, `--core`, and `--connect` as the 3 ways to "attach" for > lack of a better word. It would have been nice if the clhsdb command did a > better job of copying these three command line arguments. You are suggesting > that adding a `connect` command would have similarity to the `--connect` > option, but unfortunately the `attach` command has no such similarity with > the command line arguments `--pid` and `--core`, so I'm not so sure doing so > with `connect` is actually helping in that regard, especially when "connect" > and "attach" pretty much mean the same thing. > > As for numeric host names, yes, that is possible, but then you could also > have a numeric core file name. If we really want to avoid the numeric host > name problem, perhaps something like `attachd` would be better than > `connect`, but it seems any choice we make will have it's drawbacks due the > the baggage of existing commands and options. > @plummercj Ok, I try to add debug server support to `attach` clhsdb command. > Then should we still need CSR? Pushed new commit to be implemented as `attach`. It works fine with serviceability/sa jtreg tests on my Linux x64. I will add CSR if you are ok. ------------- PR: https://git.openjdk.java.net/jdk/pull/2773
