[
https://issues.apache.org/jira/browse/DISPATCH-1783?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17201758#comment-17201758
]
ASF GitHub Bot commented on DISPATCH-1783:
------------------------------------------
jiridanek edited a comment on pull request #853:
URL: https://github.com/apache/qpid-dispatch/pull/853#issuecomment-698563800
There is multiple options how to do this. I've been writing a more complete
list, but I only have an example for two.
## preprocessor magic
See the PR
## elfspy
https://github.com/mollismerx/elfspy
The code below, plus need to add a library
```
#include "qdr_doctest.h"
#include "elfspy/SPY.h"
#include "elfspy/Fake.h"
int fake_vsnprintf(char * s, size_t n, const char * format, va_list arg) {
return -1;
}
TEST_CASE("safe_snprintf_vsnprintf_failed") {
const int OUTPUT_SIZE = 128;
const char *TEST_MESSAGE = "something";
const int LEN = strlen(TEST_MESSAGE);
size_t len;
char output[OUTPUT_SIZE];
// weird elfspy boilerplate
char* argv[2] = {(char*)"c_unittests", nullptr};
spy::initialise(0, argv);
// setup the fake, it will be unset when the variables go out of scope
// it changes dynamic linking in the running process, to replace call to
vsnprintf in glibc with the fake
auto vsnprintf_ = SPY(&vsnprintf);
auto vsnprintf_fake = spy::fake(vsnprintf_, &fake_vsnprintf);
// alternative way for previous line:
// auto vsnprintf_fake = spy::fake(vsnprintf_, [](auto ...) { return -1;
}); // I can haz lambdaz !!!
// run the test, simulating a failed vsnprintf
output[0] = 'a';
len = safe_snprintf(output, LEN+10, TEST_MESSAGE);
CHECK(0 == len);
CHECK('\0' == output[0]);
CHECK("" == output);
}
```
## more possibilities
* use the `__wrap` feature in gnu linker
* libraries similar to elfspy, such as mimick,
https://github.com/Snaipe/Mimick
* linking order, just define vsnprintf in the same translation unit as the
test and things work out just right
* LD_PRELOAD
* ...
## general consideration
In any case, static functions cannot be mocked in C, and also inlining, when
compiler decides to optimize this way, destroys the ability to replace a
function. That implies that this kind of tests can be either only run in a
debug build, or it has to build a second, `-testing` version of the dispatch
library, for this testing. Third option is not to link to the dispatch library
and `#include` the needed .c files. That also works, if the surface area tested
is small. It encourages creating many smaller ctest test targets, instead of
one large one.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Add ability to use test doubles (mocks, fakes, ...) in C tests
> --------------------------------------------------------------
>
> Key: DISPATCH-1783
> URL: https://issues.apache.org/jira/browse/DISPATCH-1783
> Project: Qpid Dispatch
> Issue Type: Improvement
> Components: Tests
> Affects Versions: 1.14.0
> Reporter: Jiri Daněk
> Priority: Major
>
> During the work on https://github.com/apache/qpid-dispatch/pull/684, there
> was at one point a version of the PR, which tested what happens if
> {{vsnprintf}} called from {{safe_snprinf}} returns {{-1}}.
> This was done by replacing the {{vsnprintf}} with a mock (technically it's
> probably a stub, or maybe fake, but let's ignore the proper terminology).
> There are various possibilities how to approach mocking in C. They are less
> nice than in more dynamic languages such as Java, or even C++ (where it's
> possible to subclass things). Nevertheless, it can be useful technique, and
> it should be explored.
> CC [~cans]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@qpid.apache.org
For additional commands, e-mail: dev-h...@qpid.apache.org
