This is the email message from Carsten Varming:
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.
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.