RFR: JDK-8319307: DCmds should not assert on truncation and should report truncation [v2]

[David Holmes]

On Tue, 28 Nov 2023 07:18:36 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

>> src/hotspot/os/linux/os_linux.cpp line 1892:
>> 
>>> 1890:   buf[32] = '\0';
>>> 1891:   unsigned lines = 0;
>>> 1892:   while (!st->was_truncated() && (bytes = ::read(fd, buf, sizeof(buf)-1)) > 0) {
>> 
>> Okay so we stop printing if `st` is full but where is the notification that this has happened?
>
> Caller needs to take care of it.
> 
> In my mind, the caller knows what stream class it uses, so it knows if the stream is one that can truncate. So its his responsibility to handle stream::is_truncated. In this case, the caller that uses a truncatable stream is the dcmd framework, and it handles this.
> 
> The printing routines themselves could be completely oblivious to it, I only query truncated in a handful of places to save work - if we know the output stream truncated, there is no reason to continue work printing (and potentially, holding a lock while doing it).

Okay, so it isn't clear to me that all of these situations are such that the caller is always a DCmd and that it always reports the truncation due to the attach code changes.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/16474#discussion_r1407347975