wcfs: tests: Always start tDB with ZBigFile pre-created before WCFS startup
This should hopefully exercise codepaths in wcfs.go a bit more for mistakes similar to a7bf0311 (wcfs: Fix crash if on invalidation handledδZ needs to access ZODB) where the code on server side forgets to put zhead's transaction into context. Currently, because watching @tail is disallowed, this leads to panic triggered by test_wcfs_watch_setup: @at0 (03e59e3e606b89bb) -> @at1 (03e59e3e610692bb) -> @at2 (03e59e3e612a5811) -> @at3 (03e59e3e614fa9cc) -> @at4 (03e59e3e6189c3ee) -> @at5 (03e59e3e61af0baa) C: setup watch f<0000000000000002> @at0 (03e59e3e606b89bb) # pinok: {0: @at0 (03e59e3e606b89bb), 2: @at0 (03e59e3e606b89bb), 3: @at0 (03e59e3e606b89bb), 5: @at0 (03e59e3e606b89bb)} panic: at out of bounds: at: @03e59e3e606b89bb, (tail, head] = (@03e59e3e606b89bb, @03e59e3e61af0baa] goroutine 187 [running]: lab.nexedi.com/nexedi/wendelin.core/wcfs/internal/zdata.panicf(...) /home/kirr/src/neo/src/lab.nexedi.com/nexedi/wendelin.core/wcfs/internal/zdata/misc.go:47 lab.nexedi.com/nexedi/wendelin.core/wcfs/internal/zdata.(*ΔFtail).BlkRevAt(0xc000077d40, {0x969718, 0xc000062940}, 0xc0003060c0, 0x4174f4, 0x3e59e3e606b89bb) /home/kirr/src/neo/src/lab.nexedi.com/nexedi/wendelin.core/wcfs/internal/zdata/δftail.go:1077 +0xa45 main.(*WatchLink).setupWatch(0xc000108050, {0x969718, 0xc000062940}, 0x2, 0x3e59e3e606b89bb) /home/kirr/src/neo/src/lab.nexedi.com/nexedi/wendelin.core/wcfs/wcfs.go:1754 +0xe3f main.(*WatchLink)._handleWatch(0x0, {0x969718, 0xc000062940}, {0xc00001c812, 0xa00000}) /home/kirr/src/neo/src/lab.nexedi.com/nexedi/wendelin.core/wcfs/wcfs.go:1973 +0x65 main.(*WatchLink).handleWatch(0x74039b, {0x969718, 0xc000062940}, 0xc0000a4280, {0xc00001c812, 0x28}) /home/kirr/src/neo/src/lab.nexedi.com/nexedi/wendelin.core/wcfs/wcfs.go:1955 +0x10c main.(*WatchLink)._serve.func3({0x969718, 0xc000062940}) /home/kirr/src/neo/src/lab.nexedi.com/nexedi/wendelin.core/wcfs/wcfs.go:1944 +0x3c lab.nexedi.com/kirr/go123/xsync.(*WorkGroup).Go.func1() /home/kirr/src/neo/src/lab.nexedi.com/kirr/go123/xsync/xsync.go:86 +0x68 created by lab.nexedi.com/kirr/go123/xsync.(*WorkGroup).Go /home/kirr/src/neo/src/lab.nexedi.com/kirr/go123/xsync/xsync.go:83 +0x92 >>> Change history by file: f<0000000000000002>: 0 1 2 3 4 5 6 7 a b c d e f g h @at0 (03e59e3e606b89bb) @at1 (03e59e3e610692bb) 2 @at2 (03e59e3e612a5811) 2 3 4 5 @at3 (03e59e3e614fa9cc) 0 2 5 @at4 (03e59e3e6189c3ee) 2 4 5 @at5 (03e59e3e61af0baa) 3 5 However next we will anyway need to allow to setup watches @tail, and so we will be fixing this and other errors in followup commits. NOTE: we don't loose coverage for the case when ZBigFile is created after wcfs startup due to test_wcfs_watch_2files, where that scenario is tested. ΔFtail/ΔBtail tests also exercise ZBigFile/BTree epochs (creation/deletion) well.
Showing
Please register or sign in to comment