[DTrace-devel] [PATCH 2/2] test: add a test for libctf bug 30264

[Nick Alcock]

On 24 Mar 2023, Kris Van Hees via DTrace-devel spake thusly:

> On Thu, Mar 23, 2023 at 09:22:53PM -0400, Eugene Loh via DTrace-devel wrote:
>> Reviewed-by: Eugene Loh <eugene.loh at oracle.com>
>> though I confess there is stuff here I do not understand.  E.g.,
>> 
>>     $ grep rds /proc/modules
>>     $ sudo build/run-dtrace -n 'BEGIN { trace(offsetof(struct rds_message,
>> atomic)); exit(0) }'
>>     dtrace: description 'BEGIN ' matched 1 probe
>>     CPU     ID                    FUNCTION:NAME
>>       2      1                           :BEGIN         296
>> 
>> I do not need to have rds loaded?
>> 
>> (Also, the offset is nonzero, but I confess I don't know what CTF I'm
>> using.)
>
> The difference is that you are using struct rds_message whereas Nick is using
> the more specific struct rds`rds_message type specification.

I guess rds_message must have been promoted into the shared CTF. How
strange :) still we really shouldn't rely on that. This is not the only
test module-loading fixes, anyway. There are tests that rely on other
modules being loaded, some of which (like isofs) are very unlikely to be
loaded otherwise.