Let jvmtiGen exit with a non-zero exit code upon failure (original) (raw)

Carsten Varming varming at gmail.com
Fri Oct 30 00:00:28 UTC 2015


Dear Serguei and Steffan,

Thank you for the reviews. I have updated the webrev at http://cr.openjdk.java.net/~cvarming/jvmtiGen . I haven't been able to log into JBS yet, so please go ahead and create bug id for this change.

Dan: Thank you for forwarding this email to the proper alias.

Carsten

On Thu, Oct 29, 2015 at 6:44 PM, serguei.spitsyn at oracle.com < serguei.spitsyn at oracle.com> wrote:

Carsten,

I forgot to thank you for taking care about this issue! Thanks, Serguei

On 10/29/15 15:41, serguei.spitsyn at oracle.com wrote: Hi Carsten,

The fix looks good. I share the Staffan's comments though. If understand correctly, you do not have an openjdk author status yet. I will sponsor your fix. Please, let me know the bug ID after you create one.

Thanks, Serguei Please, let me know On 10/29/15 14:22, Staffan Larsen wrote: Carsten, This looks good with a few comments: 1) If you make the “verbose” variable into a static field, you can avoid the final-copying. 2) nit: Line 216: put "System.exit(1);” on it’s own line Oh, and create a bug: https://bugs.openjdk.java.net Thanks, /Staffan On 29 okt. 2015, at 14:54, Daniel D. Daugherty <_ _daniel.daugherty at oracle.com> wrote:

JVM/TI belongs to the Serviceability team so adding serviceability-dev@ ... Dan

On 10/28/15 8:45 PM, Carsten Varming wrote: webrev: http://cr.openjdk.java.net/~cvarming/jvmtiGen/ bug: ? jvmtiGen is used to process a number of xml and xslt files in OpenJDK. Currently jvmtiGen exits with exit code 0 regardless of its success. This causes make to often consider a target finished when in fact the target failed. It also leads to funny error checking after the execution of jvmtiGen. For instance, in many trace.make files[*] a test for the existence of the output file is carried out after the completion of jvmtiGen. In a clean working repository that test is equivalent to jvmtiGen exiting with a proper exit failure code on failure, but in a dirty working repository the target file might just be pre-existing. This causes unnecessary pain when working with files processed by jvmtiGen. In this change I chose to exit with exit code 1 whenever a failure is detected, be it a dtd validation failure, an IO failure, or something else entirely. This halts the building of OpenJDK on failures and ultimately makes development easier. I also added a verbose option such that warnings from the xml parser and dtd checker can be printed on stderr if desired. Finally, I changed all the error message printing to stderr. :-) Let me know what you think. BTW. This is the first time I tried the webrev system, so hopefully it all looks good. I havn't figured out how to create a bug yet, whence the question mark. I wasn't sure if hotspot-runtime-dev is the right email alias. Please let me know if there is a more appropriate alias for this email. [*] Why are so many of the non-shared makefiles almost identical?



More information about the hotspot-runtime-dev mailing list