Cleanup files uploaded

[LeSpocky]

Requirements for Contributing a Bug Fix

• Fill out the template below. Any pull request that does not include enough information to be reviewed in a timely manner may be closed at the maintainers' discretion.
• The pull request must only fix an existing bug. To contribute other changes, you must use a different template. You can see all templates at https://github.com/etr/libhttpserver/tree/master/.github/PULL_REQUEST_TEMPLATE.
• The pull request must update the test suite to demonstrate the changed functionality.
• After you create the pull request, all status checks must be pass before a maintainer reviews your contribution. For more details, please see https://github.com/etr/libhttpserver/tree/master/CONTRIBUTING.md#pull-requests.

Identify the Bug

#264

Description of the Change

The list of files lives in the httpserver::http_request object. The destructor is called after the request handler is completed, so we can assume user had done all tasks needed for this request. Simply cleanup those files, before the files map is gone.

Alternate Designs

none considered

This is the most simple approach.

Possible Drawbacks

If the user needs those files, she should copy/move theme in the request
handler.

Verification Process

Integration tests in my application, not public.

Release Notes

"N/A" because follow-up to #257.

[LeSpocky]

Throwing this in as a draft, to see if this goes in the right direction.

1. Not entirely sure about that friend class thing. I suppose the destructor should be private, because the constructors are also private?
2. Need to come up with some way to test this in libhttpserver integration tests, right?

[LeSpocky]

Modified the integration tests to check on deleted files. Hope this is sufficient to show the mechanism works as it should.

[LeSpocky]

Modified the integration tests to check on deleted files. Hope this is sufficient to show the mechanism works as it should.

On some machines those tests fail (and sometimes with a strange 0!=0 result) and on some they don't.

Can this happen because the webserver runs in a different thread than the test and this is some kind of concurrency problem?

[etr]

Hey,

Thanks for the change.

Not entirely sure about that friend class thing. I suppose the destructor should be private, because the constructors are also private?

Yes, it is fine with the destructor being private and using friends to access it.

Need to come up with some way to test this in libhttpserver integration tests, right?

I guess you have the test now or is there something else you were thinking?

Can this happen because the webserver runs in a different thread than the test and this is some kind of concurrency problem?

It might be that the check is happening before the deletion has happened. Is this still happening or your latest change has fixed it?

[LeSpocky]

Thanks for looking into my changes. Answers to your questions below.

Need to come up with some way to test this in libhttpserver integration tests, right?

I guess you have the test now or is there something else you were thinking?

Correct. I added the changes to the tests few days after the work in the destructor.

Can this happen because the webserver runs in a different thread than the test and this is some kind of concurrency problem?

It might be that the check is happening before the deletion has happened. Is this still happening or your latest change has fixed it?

No, the last change to the tests fixed that. CI did complain before, but not anymore after that.

[etr]

Thanks again for the help. All approved.