宋文武 <iyzs...@gmail.com> skribis: > Ludovic Courtès <l...@gnu.org> writes: > >> 宋文武 <iyzs...@gmail.com> skribis: >> >>> * gnu/packages/database.scm (mariadb): New variable. >> >> +s: “databases.scm” >> >>> + 'configure 'pre-configure >>> + (lambda _ >>> + (setenv "CONFIG_SHELL" (which "sh")) >>> + ;; XXX: libstdc++.so lacks RUNPATH for libgcc_s.so. >>> + (setenv "LDFLAGS" "-lgcc_s") >> >> Indeed, but this is normally compensated by the fact that our GCC >> automatically passed -Wl,-rpath=$gcclibdir, no? What error where you >> getting without it? > Without this flag, all binaries are broken (they linked with > libstdc++.so but with libgcc_s.so not found).
On master or core-updates? How is that possible? All binaries have GCC’s lib output in their RUNPATH AFAIK. >>> + ;; Install testing data to the 'test' output. >>> + (mkdir-p test) >>> + (system* "mv" (string-append out "/data") test) >>> + (system* "mv" (string-append out "/mysql-test") test) >>> + (system* "mv" (string-append out "/sql-bench") test))) >> >> Could you use ‘rename-file’? > It seem that those are assumed to be installed with the same > basedir as mariadb, and I failed to trick it to run. > > I'd like to remove those test data (and the 'test' output). Actually yes, why is this “test” output needed in the first place? Ludo’.
