patchrefs: print errno for read_raw_ref if GIT_TRACE_REFS is set

From: Han-Wen Nienhuys <hanwen@google.com>

The ref backend API uses errno as a sideband error channel.

Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
---
    refs: print errno for read_raw_ref if GIT_TRACE_REFS is set
    
    The ref backend API uses errno as a sideband error channel.
    
    Signed-off-by: Han-Wen Nienhuys hanwen@google.com

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1002%2Fhanwen%2Ferrno-debug-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1002/hanwen/errno-debug-v1
Pull-Request: https://github.com/git/git/pull/1002

 refs/debug.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

"Han-Wen Nienhuys via GitGitGadget" <gitgitgadget@gmail.com> writes:

OK. Between trace_printf_key() and the return of the call to read_raw_ref() method of the ref backend is only an "if (res == 0)" condition and I do not see anything that might clobber errno.

I do wonder if we want strerror(errno) instead of the number, but I can go either way (it's just a trace output).

Thanks, will queue.

On Mon, Apr 12, 2021 at 11:45 PM Junio C Hamano <gitster@pobox.com> wrote:

I don't want to bet on that. Let me send a second round.

For tracing, it would be most useful if we got the actual symbol (eg. ENOENT). Is there a way to get at that?

On Tue, Apr 13, 2021 at 1:58 PM Han-Wen Nienhuys <hanwen@google.com> wrote:

oh, ugh.  In email, there are two calls, but one is prefixed with '-'.

Nevertheless, a bit of paranoia doesn't hurt here.

Han-Wen Nienhuys <hanwen@google.com> writes:

I do not think there is.

And that is the reason why we see everywhere calls to die_errno(), error(..., strerror(errno)), etc.

As this is interpolated into trace with untranslated string, I suspect that strerror() is not a good match, so let's live with %d for now.

Thanks.

On Tue, Apr 13, 2021 at 10:08 PM Junio C Hamano <gitster@pobox.com> wrote:

Great!

This topic is marked as

Waiting for reviews to conclude. cf. <xmqq4kgbb2ic.fsf@gitster.g>

but I don't know what is left to do. Oversight?

From: Han-Wen Nienhuys <hanwen@google.com>

The ref backend API uses errno as a sideband error channel.

Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
---
    refs: print errno for read_raw_ref if GIT_TRACE_REFS is set
    
    The ref backend API uses errno as a sideband error channel.
    
    Signed-off-by: Han-Wen Nienhuys hanwen@google.com

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1002%2Fhanwen%2Ferrno-debug-v2
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1002/hanwen/errno-debug-v2
Pull-Request: https://github.com/git/git/pull/1002

Range-diff vs v1:

 1:  9b150c5563f9 ! 1:  1e9a8990e7f2 refs: print errno for read_raw_ref if GIT_TRACE_REFS is set
     @@ Commit message
      
       ## refs/debug.c ##
      @@ refs/debug.c: static int debug_read_raw_ref(struct ref_store *ref_store, const char *refname,
     + {
     + 	struct debug_ref_store *drefs = (struct debug_ref_store *)ref_store;
       	int res = 0;
     ++	int saved_errno = 0;
       
       	oidcpy(oid, &null_oid);
      +	errno = 0;
       	res = drefs->refs->be->read_raw_ref(drefs->refs, refname, oid, referent,
       					    type);
     ++	saved_errno = errno;
       
     -@@ refs/debug.c: static int debug_read_raw_ref(struct ref_store *ref_store, const char *refname,
     + 	if (res == 0) {
       		trace_printf_key(&trace_refs, "read_raw_ref: %s: %s (=> %s) type %x: %d\n",
       			refname, oid_to_hex(oid), referent->buf, *type, res);
       	} else {
      -		trace_printf_key(&trace_refs, "read_raw_ref: %s: %d\n", refname, res);
      +		trace_printf_key(&trace_refs,
      +				 "read_raw_ref: %s: %d (errno %d)\n", refname,
     -+				 res, errno);
     ++				 res, saved_errno);
       	}
       	return res;
       }

 refs/debug.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)