On Wed, Jan 10, 2018 at 7:43 AM, Rainer Orth <r...@cebitec.uni-bielefeld.de> wrote: > Hi Ian, > >> On Wed, Jan 10, 2018 at 3:44 AM, Rainer Orth >> <r...@cebitec.uni-bielefeld.de> wrote: >>> >>> thanks. Testing has now concluded as well. x86 results are good (no >>> regressions except for cmd/internal/buildid which fails on Linux, too), >>> as are 64-bit sparc results. >> >> The cmd/internal/buildid test does pass on my system. What are you seeing? > > the log shows > > --- FAIL: TestReadFile (0.01s) > buildid_test.go:40: ReadFile(testdata/p.a) = "", open testdata/p.a: > no such file or directory, want > "abcdefghijklmnopqrstuvwxyz.1234567890123456789012345678901234567890123456789012345678901234", > nil > buildid_test.go:47: ReadFile(testdata/p.a) [readSize=2k] = "", open > testdata/p.a: no such file or directory, want > "abcdefghijklmnopqrstuvwxyz.1234567890123456789012345678901234567890123456789012345678901234", > nil > buildid_test.go:52: open testdata/p.a: no such file or directory > > libgo/go/cmd/internal/buildid/testdata/p.a is just missing.
Ah. Turns out it was being ignored by the default global-ignores setting in the Subversion configuration, which ignores all .a files. I found a couple of other .a files that were missing in the gofrontend git repo (dropped by merge.sh) and copied them over. Then I committed all the missing .a files to the gcc Subversion repository (there were five missing files; the others were not causing a test failure). Ian
