[pacman-dev] [PATCH] Print callback messages to stderr

[Dave Reisner]

On Sat, Aug 20, 2011 at 10:11:11AM -0500, Dan McGee wrote:
> On Sat, Aug 20, 2011 at 7:58 AM, Dave Reisner <d at falconindy.com> 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<allan at archlinux.org>
> >> >---
> >> >  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
> 

Probably because we only timestamp when we build with --enable-debug so
its #ifdef'd. That said, I don't think the timestamps are incredibly
intrusive and they add value to debug output when we request it from
users. We could kill two birds with one stone here and get rid of
the #ifdef to just print the log line all at once.

d