[
https://issues.apache.org/jira/browse/THRIFT-857?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12900354#action_12900354
]
Bruce Lowekamp commented on THRIFT-857:
---------------------------------------
I submitted a THRIFT-859 (with patch) that proposes reverting the check that
caused this problem in the first place.
t_program.h was revised to include (everything but the last
namespace[language]=namespace assignment:
void set_namespace(std::string language, std::string name_space) {
t_generator_registry::gen_map_t my_copy =
t_generator_registry::get_generator_map();
t_generator_registry::gen_map_t::iterator it;
it=my_copy.find(language);
if (it == my_copy.end()) {
if (language != "smalltalk.prefix" && language != "smalltalk.package" &&
language != "py.twisted") {
throw "No generator named '" + language + "' could be found!";
}
}
namespaces_[language] = name_space;
}
Not only does this blow up when a generator is disabled, but it also
- prevents anyone from adding a new sub-namespace without modifying this code,
as with smalltalk.prefix and smalltalk.package
- actually skips the check it's intended to do in the smalltalk case, i.e. you
can declare a smalltalk.prefix namespace without having the smalltalk generator
enabled
We ran into this because we had a mod to declare a separate namespace for
py.twisted, which is what I'm about to put an issue in for.
Now, if it's really important to check that the generator is there for a
namespace declaration, then the *right* way to do this is to take the base
language name, check that there's a generator for it, then if there's a
sub-language (smalltalk.package or py.twisted) call into the generator to see
if it recognizes the declaration.
That seems like a lot of effort to put the support into every generator for
relatively little benefit. I care a lot more about whether there's a missing
namespace declaration for a language I'm trying to generate than if there's a
namespace declaration for a language I'm not trying to generate.
So as part of THRIFT-859, I think we should just revert the change to
t_program.h, allow any namespace declaration to be made, and let the generator
report an error if it's unhappy with there not being a namespace, etc.
> tests run by "make install" fail if generators are disabled
> -----------------------------------------------------------
>
> Key: THRIFT-857
> URL: https://issues.apache.org/jira/browse/THRIFT-857
> Project: Thrift
> Issue Type: Bug
> Components: Test Suite
> Affects Versions: 0.4
> Environment: mac os
> Reporter: Bruce Lowekamp
> Priority: Minor
>
> configuring with generators disabled (configure --disable-gen-java, for
> example) produces a failed build because the tests run by "make install"
> require the java, cpp, rb, perl, csharp, and js generators.
> I would personally favor either removing those --disble-gen options
> altogether, or adding a warning message in the configure --help output that
> disabling those generators is not recommended/will cause test failure.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
