On 2015-08-23 16:50, Niall Douglas wrote:
On 23 Aug 2015 at 9:37, Roland Bock wrote:
Monday on July-20. But the [new] review start date was Friday August-21. I delivered a review ready AFIO on Friday. Ahmed is currently busy, and I am sure once he becomes unbusy he will announce the review in the usual way. Out of curiosity:
What's the big rush? Just a month ago, the library was in sucha state, that you canceled the review. Now you just "finished" a ready-for-review library on the day the review was supposed to start? There is no rush. The library was finished in its present form around August 8th (you can check the commit log). Since then it's been almost entirely dotting i's and crossing t's and writing the new tutorial.
And the review is supposed to have started even though the review manager was so busy that he could not even send a single mail. We all have unexpected things turn up when we don't expect. Ahmed is doing his best, same as all of us.
You also mentioned:
A later *internal* implementation will be completely new, but that isn't important to this review because little external changes. To me this is another indicator that maybe it is a bit early.
One of the aspects of the review is to look at the implementation. Please DO look at the implementation. Any code which calls a filing system API won't be changing. The stuff which will be (internally) changing is how the ASIO reactor is used. That's all.
Sounds like we can drop that item for this review, since it won't last too long anyway? The four APIs scheduled for deprecation listed at https://boostgsoc13.github.io/boost.afio/doc/html/afio/release_notes.h tml can be skipped. Otherwise all other implementation is up for review.
Don't get me wrong, maybe you are just the victim of your own honesty and full disclosure policy (admirable traits, btw). But judging by the mails you wrote in the last few days, if it were my library, I would kindly ask to treat this more like a pre-review of the new design. The full review would then be done once the implementation is as you want it. I think some are minded to blow out of proportion how "unfinished" AFIO is for various personal reasons or out of ignorance.
I would not be submitting AFIO for review here if I did not consider it production ready.
And I should emphasise that the choice to implement the new API on top of the mature engine was something I came here to boost-dev to ask for advice upon many months ago, and what you have here today is what was wanted then. I have delivered what was asked of me - admittedly a month late, but three weeks of that went on writing a new tutorial which was asked of me by people here on boost-dev last month. If I hadn't have written the new tutorial due to the feedback last month, I would have been just a week late, not bad for a ~400 man hour project.
I am happy to continue to receive feedback from people here on what further work needs to be done, and implement solutions to problems found by reviewers as I have been doing since 2013. That's why the library is up for review.
Thanks for your answer! I will try to take a closer look, although I cannot promise to get a review done in time (lots on my table these days). Roland