RFR: 7164841: Improvements to the GC log file rotation

[Yumin Qi]

Thomas,

   This is the new version:
http://cr.openjdk.java.net/~minqi/7164841/webrev05 
<http://cr.openjdk.java.net/%7Eminqi/7164841/webrev05>

   The change is adding internal tests after make_log_name_internal 
added to implement the original make_log_name function.

   Please let me know if it is OK. JPRT task is under building/testing, 
so far it is good.

Thanks
Yumin

On 9/12/2013 8:24 AM, Thomas Schatzl wrote:
> Hi Yumin,
>
> On Wed, 2013-09-11 at 15:45 -0700, Yumin Qi wrote:
>> Thomas,
>>
>>     Attached are two files, on is script file for running and the other
>> is the output result.
>>     With this flag
>>
>> ExecuteInternalVMTests
>>
>> on, the run only print out so much, and did not contine to do rest of
>> the test. Think it is right.
>    sorry for being unclear: the request has been to add new tests to the
> internal VM tests, not only run them. The method to replace the %p/%t
> markers would make them an ideal candidate for such testing.
>
> E.g. have a method that compares output of that function with some known
> values.
>
> However I did not think it through completely, as the values of %p/%t
> are dependent on external circumstances, and it is maybe not worth to
> add infrastructure to allow fake values (well, one could extract the
> generation of these values out, and pass them to the function under
> test).
>
> I.e.  change make_log_name()
>
> to this signature:
>
> make_log_name_internal(const char* log_name, const char*
> force_directory, const char* datetime, int pid) {
>
>    ... use datetime, pid in the code instead of retrieving them here...
> }
>
> make_log_name(const char* log_name, const char* force_directory) {
>    char tms[32];
>    get_datetime_string(tms, sizeof(tms);
>
>    make_log_name_internal(log_name, force_directory, tms,
> os::current_process_id());
> }
>
> then you could easily write tests like:
>
>    char* fixed_time = "1980-20-....";
>    char* result;
>
>    result = make_log_name_internal("somename%p", false, fixed_time, 200);
>    if (strncmp(result, "somename200") != 0) { // error out }
>    free result
>
>    [... more tests...]
>
> and so on. I am not sure how you did the testing while developing, but
> the effort would probably only be a matter of converting them to this
> format.
>
> Thomas
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/attachments/20130912/71dca92a/attachment.html