On Sat, Aug 20, 2011 at 7:58 AM, Dave Reisner <[email protected]> wrote: > On Sat, Aug 20, 2011 at 03:15:34PM +1000, Allan McRae wrote: >> On 20/08/11 12:42, Allan McRae wrote: >> >Fixes FS#25099. >> > >> >Signed-off-by: Allan McRae<[email protected]> >> >--- >> > src/pacman/callback.c | 2 +- >> > 1 files changed, 1 insertions(+), 1 deletions(-) >> > >> >diff --git a/src/pacman/callback.c b/src/pacman/callback.c >> >index 873e3fc..5ee4e5a 100644 >> >--- a/src/pacman/callback.c >> >+++ b/src/pacman/callback.c >> >@@ -690,7 +690,7 @@ void cb_log(alpm_loglevel_t level, const char *fmt, >> >va_list args) >> > output = alpm_list_add(output, string); >> > } >> > } else { >> >- pm_vfprintf(stdout, level, fmt, args); >> >+ pm_vfprintf(stderr, level, fmt, args); >> > } >> > } >> > >> >> >> This breaks some pactests because stdout/stderr output is not being >> kept in sync so timestamps with --debug get printed all over the >> place. e.g. >> >> > ./src/pacman/pacman -T glibc --debug >> <snip> >> debug: unregistering database 'local' >> debug: freeing package cache for repository 'local' >> debug: unregistering database 'allanbrokeit' >> debug: unregistering database 'kernel64' >> debug: unregistering database 'testing' >> debug: unregistering database 'core' >> debug: unregistering database 'extra' >> debug: unregistering database 'community-testing' >> debug: unregistering database 'community' >> [15:11:34] [15:11:34] [15:11:34] [15:11:34] [15:11:34] [15:11:34] >> [15:11:34] [15:11:34] [15:11:34] >> >> > > Wouldn't it make more sense to write the timestamps to stderr along with > the logging they belong to? If someone were to enable --debug and send > stderr off to a file, you'd see some really weird output.
Yes- we shouldn't be trying to multiplex stderr and stdout correctly on the same line. These need to both go to the same place, and why aren't they printed in the same printf() invocation? -Dan
