Skip to contentSkip to conversation

xfs_db: redirect printfs when metadumping to stdout

PATCH Superseded Darrick J. Wong 2 participants 2 messages Jul 14, 2017 – Jul 15, 2017

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xfs_db: redirect printfs when metadumping to stdout

If we're metadumping to stdout, we don't want xfs_db's various dbprintf statements dumping to stdout because that'll corrupt the metadump. Therefore, let outf point to the existing stdout and redirect stdout to stderr for the duration of the dump operation.

---
 db/metadump.c |    9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
db/metadump.c +7 -2
index e046f60..e6e153d 100644
@@ -2874,6 +2874,7 @@ metadump_f(
 	xfs_agnumber_t	agno;
 	int		c;
 	int		start_iocur_sp;
+	bool		stdout_metadump = false;
 	char		*p;
 
 	exitcode = 1;
@@ -2989,6 +2990,8 @@ metadump_f(
 			return 0;
 		}
 		outf = stdout;
+		stdout = stderr;
+		stdout_metadump = true;
 	} else {
 		outf = fopen(argv[optind], "wb");
 		if (outf == NULL) {
@@ -3020,9 +3023,11 @@ metadump_f(
 		exitcode = write_index() < 0;
 
 	if (progress_since_warning)
-		fputc('\n', (outf == stdout) ? stderr : stdout);
+		fputc('\n', stdout_metadump ? stderr : stdout);
 
-	if (outf != stdout)
+	if (stdout_metadump)
+		stdout = outf;
+	else
 		fclose(outf);
 
 	/* cleanup iocur stack */
#1 lore raw
Quoted 24 lines · Darrick J. Wong

If we're metadumping to stdout, we don't want xfs_db's various dbprintf statements dumping to stdout because that'll corrupt the metadump. Therefore, let outf point to the existing stdout and redirect stdout to stderr for the duration of the dump operation.

---
 db/metadump.c |    9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
db/metadump.c +1 -0
index e046f60..e6e153d 100644
@@ -2874,6 +2874,7 @@ metadump_f(
 	xfs_agnumber_t	agno;
 	int		c;
 	int		start_iocur_sp;
+	bool		stdout_metadump = false;
 	char		*p;
 
 	exitcode = 1;

db/metadump.c metadump_f( · line 2994

Quoted 5 lines
db/metadump.c +2 -0
@@ -2989,6 +2990,8 @@ metadump_f(
 			return 0;
 		}
 		outf = stdout;
+		stdout = stderr;
+		stdout_metadump = true;

Ok, I ... guess? xfs_metadump already takes pains to print to stderr as appropriate; see print_warning(), and comments about where "wornings" should go.

But in this case I guess it's the guts of db / dbprintf that caused the problems, right? ... yup you said that in the commit message.

I suppose a comment:

/* Move stdout to stderr so thta xfs_db warnings don't pollute metadump stream */

or something would help the future reader know why you had to do this.

it feels a little hacky but I don't really see a particularly better way I guess.

-Eric

Quoted 18 lines
db/metadump.c +4 -2
 	} else {
 		outf = fopen(argv[optind], "wb");
 		if (outf == NULL) {
@@ -3020,9 +3023,11 @@ metadump_f(
 		exitcode = write_index() < 0;
 
 	if (progress_since_warning)
-		fputc('\n', (outf == stdout) ? stderr : stdout);
+		fputc('\n', stdout_metadump ? stderr : stdout);
 
-	if (outf != stdout)
+	if (stdout_metadump)
+		stdout = outf;
+	else
 		fclose(outf);
 
 	/* cleanup iocur stack */