This isn’t it, because we always call through to swift_release in CFRelease on Linux, because the CFString is a swift type under the hood.
We may be able to change this to check for a CF specific hack of some kind, but in the meantime adding the retain works. PR here: https://github.com/apple/swift-corelibs-foundation/pull/1351 - Tony > On Nov 29, 2017, at 8:31 PM, Philippe Hausler <phaus...@apple.com> wrote: > > I think that perhaps we have a problem with the retain count flags in the > definition of CFSTR. Previously the pinned reference would prevent the > deallocation. This may have gotten clobbered with the latest CF import. > > Sent from my iPhone > > On Nov 29, 2017, at 7:55 PM, Brandon Williams via swift-corelibs-dev > <swift-corelibs-dev@swift.org <mailto:swift-corelibs-dev@swift.org>> wrote: > >> Hey Tony, >> >> We were able to fix our crashes by avoiding our uses of `URLQueryItem`s. In >> one place I had to completely remove the use of `URLQueryItem`: >> >> https://github.com/pointfreeco/swift-web/pull/79/files#diff-293b73a926ba418bd9511ef735fc947cL147 >> >> <https://github.com/pointfreeco/swift-web/pull/79/files#diff-293b73a926ba418bd9511ef735fc947cL147> >> >> And in another my colleague Stephen Celis discovered that duplicate keys in >> a query string can cause this to crash: >> >> https://github.com/pointfreeco/swift-web/pull/80 >> <https://github.com/pointfreeco/swift-web/pull/80> >> >> He was able to come up with a minimal test case to demonstrate this: >> >> https://github.com/pointfreeco/swift-web/commit/cfa9b519a829f1c1627620521326f0877c2d023a >> >> <https://github.com/pointfreeco/swift-web/commit/cfa9b519a829f1c1627620521326f0877c2d023a> >> >> And you can see the failure in this Travis CI report: >> >> https://travis-ci.org/pointfreeco/swift-web/builds/309316418 >> <https://travis-ci.org/pointfreeco/swift-web/builds/309316418?utm_source=email&utm_medium=notification> >> >> At the bottom you’ll find: >> >> ```fatal error: Constant strings cannot be deallocated: file >> Foundation/NSCFString.swift, line 118``` >> >> So definitely seems to be in that area! >> >> Now, as far as rebuilding swift-corelibs-foundation, I’m down to try but I >> dont really know much about how to do that. With some instructions I could >> give it a shot. >> >> Thanks for the help! >> >> >> >> >> On Wed, Nov 29, 2017 at 5:59 PM Tony Parker <anthony.par...@apple.com >> <mailto:anthony.par...@apple.com>> wrote: >> I see a couple of places that are suspicious there: >> >> Description: >> >> https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L66 >> >> <https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L66> >> >> And the nameString / valueString constants in the copy query items function: >> >> https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1057 >> >> <https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1057> >> https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1079 >> >> <https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1079> >> https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1104 >> >> <https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1104> >> https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1134 >> >> <https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1134> >> https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1158 >> >> <https://github.com/apple/swift-corelibs-foundation/blob/master/CoreFoundation/URL.subproj/CFURLComponents.c#L1158> >> >> Do you have the ability to re-build swift-corelibs-foundation and run a >> test? I suspect if you replace those with CFRetain(CFSTR(…)) then it will >> fix your crash. >> >> - Tony >> >> >>> On Nov 29, 2017, at 2:06 PM, Brandon Williams <mbw...@gmail.com >>> <mailto:mbw...@gmail.com>> wrote: >>> >>> Thanks for the info! >>> >>> FWIW, most of the places I have encountered this so far have made use of >>> `URLComponents`, and have even been able to eliminate the crash by getting >>> rid of that code in a few places. The other JIRA bug on this topic also >>> mentions URLComponents in their repro case. Seems to be the culprit. >>> >>> Also worth noting that in the above cases dealing with URLComponents the >>> crash only happens in DEBUG compilations, not RELEASE. >>> >>> However, I do have another one of these crashes that _does_ happen on >>> RELEASE builds that I haven’t yet been able to reduce. >>> >>> >>> On Wed, Nov 29, 2017 at 4:43 PM Tony Parker <anthony.par...@apple.com >>> <mailto:anthony.par...@apple.com>> wrote: >>> Hi Brandon, >>> >>> This is probably a bug in the CoreFoundation C sources part of >>> swift-corelibs-foundation. Unlike Darwin platforms, we can’t make the >>> CFSTR(“”) macro produce a CFStringRef that cannot be overreleased. Probably >>> there is a constant string returned from CF function, then the Swift >>> runtime assumes it can release it, and bam we wind up with an overrelease. >>> >>> We’ll have to track down which function is getting called, which probably >>> just requires stepping through that test case with a debugger… >>> >>> - Tony >>>> On Nov 29, 2017, at 1:34 PM, Brandon Williams via swift-corelibs-dev >>>> <swift-corelibs-dev@swift.org <mailto:swift-corelibs-dev@swift.org>> wrote: >>>> >>> >>>> Hello all! >>>> >>>> We’ve been encountering this runtime error quite a bit recently, and we >>>> have no idea why. We’ve filed a bug and there’s one other on JIRA related >>>> to this, but both without any comments: >>>> >>>> https://bugs.swift.org/browse/SR-6422 >>>> <https://bugs.swift.org/browse/SR-6422> >>>> https://bugs.swift.org/browse/SR-6398 >>>> <https://bugs.swift.org/browse/SR-6398> >>>> >>>> We’re curious if others are aware of this and or has more info? Is it a >>>> red herring for some other problem? >>>> >>>> Any info would be helpful! >>> >>>> _______________________________________________ >>>> swift-corelibs-dev mailing list >>>> swift-corelibs-dev@swift.org <mailto:swift-corelibs-dev@swift.org> >>>> https://lists.swift.org/mailman/listinfo/swift-corelibs-dev >>>> <https://lists.swift.org/mailman/listinfo/swift-corelibs-dev> >> >> _______________________________________________ >> swift-corelibs-dev mailing list >> swift-corelibs-dev@swift.org <mailto:swift-corelibs-dev@swift.org> >> https://lists.swift.org/mailman/listinfo/swift-corelibs-dev >> <https://lists.swift.org/mailman/listinfo/swift-corelibs-dev>
_______________________________________________ swift-corelibs-dev mailing list swift-corelibs-dev@swift.org https://lists.swift.org/mailman/listinfo/swift-corelibs-dev
