<div dir="auto">Actually, I already have a fork on which I managed to get this info into Checkers.td (and have come to the same conclusion about tblgen being difficult to manage).<br><div class="gmail_extra" dir="auto"><br></div><div class="gmail_extra" dir="auto">Hashtags sound awesome, but I'll probably be burnt out with options by the time I finish the already planned changes.</div><div class="gmail_extra" dir="auto"><br><div class="gmail_quote">On 29 Oct 2018 19:39, "Artem Dergachev" <<a href="mailto:noqnoqneo@gmail.com">noqnoqneo@gmail.com</a>> wrote:<br type="attribution"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">What do you think about refactoring Checkers.td into a .def-file and listing checker options there? Eg.,<br>
<br>
CHECKER(Malloc, core,<br>
"Check for memory leaks, double free, and use-after-free problems.")<br>
OPTION(Malloc, Optimistic, bool, false,<br>
"A useless option that needs to be removed.")<br>
<br>
CHECKER(PthreadLock, alpha.unix,<br>
"Simple lock -> unlock checker"<br>
<br>
We could also de-duplicate packages (though i don't think that's necessary, as it's a matter of simple string prefix comparison):<br>
<br>
BEGIN_PACKAGE(unix, alpha)<br>
CHECKER(PthreadLock, "Simple lock -> unlock checker")<br>
CHECKER(...)<br>
OPTION(...)<br>
END_PACKAGE<br>
<br>
=====<br>
<br>
As an unrelated note, i've been dreaming for a while now about replacing packages with hashtags for more flexibility. Eg.,<br>
<br>
CHECKER(PthreadLock, "Simple lock -> unlock checker", "#alpha #posix #pathsensitive #threading")<br>
<br>
Of course, we can always keep packages around for backwards compatibility.<div class="quoted-text"><br>
<br>
<br>
On 10/26/18 3:17 PM, Kristóf Umann wrote:<br>
</div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="quoted-text">
Too bad I cant edit mails.<br>
<br>
Where I talked about extracting all isn't easily accessable fields and related methods to CheckerManagerData, I actually meant *easily* accessible (since some checkers actually need to access LangOptions, as well as AnalyerOptions, both avaible when other similar -help options are handled).<br>
<br></div><div class="elided-text">
On 26 Oct 2018 20:58, "Kristóf Umann" <<a href="mailto:dkszelethus@gmail.com" target="_blank">dkszelethus@gmail.com</a> <mailto:<a href="mailto:dkszelethus@gmail.com" target="_blank">dkszelethus@gmail.com</a>><wbr>> wrote:<br>
<br>
Hi!<br>
<br>
We did have a thread about this but with a very misleading title,<br>
so here's a link to that, and I'll get into this mail:<br>
<a href="http://lists.llvm.org/pipermail/cfe-dev/2018-October/059664.html" rel="noreferrer" target="_blank">http://lists.llvm.org/pipermai<wbr>l/cfe-dev/2018-October/059664.<wbr>html</a><br>
<<a href="http://lists.llvm.org/pipermail/cfe-dev/2018-October/059664.html" rel="noreferrer" target="_blank">http://lists.llvm.org/piperma<wbr>il/cfe-dev/2018-October/<wbr>059664.html</a>><br>
<br>
AnalyzerOptions shouldn't be mutable once fully initialized (which<br>
should be achieveble by the time the actual analysis begins), and<br>
the greatest enemies of this idea are checker options, because<br>
* we can either forget about collecting all checker options and<br>
possibly diagnose them, and let checkers use AnalyzerOptions as a<br>
sort of set of user supplied options. This is the current state of<br>
things, and should my non-checker option refactoring effort go<br>
through, AnalyzerOptions can be made const straight away.<br>
* we could supply a mutable AnalyzerOptions object to the checkers<br>
when registering, let them register and evaluate their options,<br>
and make it immutable for the rest of the analysis.<br>
<br>
I'm highly in favour of the second option, but it's a non-trivial<br>
issue, mostly because of external plugins, which is why I'm<br>
looking for some feedback on my ideas.<br>
<br>
In order to register (and, more importantly, initialize) checkers,<br>
one needs to have access to a CheckerManager object, which isn't<br>
trivial to create, which makes it impossible to implement a help<br>
flag (like -analyzer-checker-help or the proposed<br>
-analyzer-config-help). I'm proposing two possible solutions.<br>
<br>
1. Extract everything that isn't easily accessible to a new<br>
CheckerManagerData class, make CheckerManager only responsible for<br>
interacting (but not registering) checkers. I've got a fork on<br>
which I managed to get this working, but I disliked this approach,<br>
and went on to find a better solution.<br>
<br>
2. Force checkers to properly register their options in a new,<br>
registerOptionsFor##CHECKERNAM<wbr>E function, which would take<br>
AnalyzerOptions as a parameter, alongside register##CHECKERNAME.<br>
This would add one more complication to the already<br>
very-not-trivial registering process, but could also be<br>
autogenerated using tblgen.<br>
<br>
It's clear to me that the second option is superior to the second,<br>
but going forward with either is a lot of work, so I'm looking for<br>
feedback.<br>
<br>
Thanks to everyone who already took the time to help me with this<br>
effort!<br>
<br>
Cheers,<br>
Kristóf<br>
<br>
</div></blockquote>
<br>
</blockquote></div><br></div></div>