2024-03-05 12:20:40

by Ben Zong-You Xie

Subject: [PATCH 1/2] perf daemon: Fix the warning about time_t

In the 32-bit platform, the size of time_t is still 64 bits. Thus, use
PRIu64 to resolve the format problem.

Signed-off-by: Ben Zong-You Xie <[email protected]>
---
tools/perf/builtin-daemon.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/tools/perf/builtin-daemon.c b/tools/perf/builtin-daemon.c
index 83954af36753..0b6ffd51c475 100644
--- a/tools/perf/builtin-daemon.c
+++ b/tools/perf/builtin-daemon.c
@@ -23,6 +23,7 @@
#include <sys/signalfd.h>
#include <sys/wait.h>
#include <poll.h>
+#include <inttypes.h>
#include "builtin.h"
#include "perf.h"
#include "debug.h"
@@ -688,7 +689,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
/* lock */
csv_sep, daemon->base, "lock");

- fprintf(out, "%c%lu",
+ fprintf(out, "%c%" PRIu64 "",
/* session up time */
csv_sep, (curr - daemon->start) / 60);

@@ -700,7 +701,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
daemon->base, SESSION_OUTPUT);
fprintf(out, " lock: %s/lock\n",
daemon->base);
- fprintf(out, " up: %lu minutes\n",
+ fprintf(out, " up: %" PRIu64 " minutes\n",
(curr - daemon->start) / 60);
}
}
@@ -727,7 +728,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
/* session ack */
csv_sep, session->base, SESSION_ACK);

- fprintf(out, "%c%lu",
+ fprintf(out, "%c%" PRIu64 "",
/* session up time */
csv_sep, (curr - session->start) / 60);

@@ -745,7 +746,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
session->base, SESSION_CONTROL);
fprintf(out, " ack: %s/%s\n",
session->base, SESSION_ACK);
- fprintf(out, " up: %lu minutes\n",
+ fprintf(out, " up: %" PRIu64 " minutes\n",
(curr - session->start) / 60);
}
}
--
2.34.1

2024-04-26 17:32:51

by Palmer Dabbelt

Subject: Re: [PATCH 1/2] perf daemon: Fix the warning about time_t

On Tue, 05 Mar 2024 04:05:00 PST (-0800), [email protected] wrote:
> In the 32-bit platform, the size of time_t is still 64 bits. Thus, use
> PRIu64 to resolve the format problem.
>
> Signed-off-by: Ben Zong-You Xie <[email protected]>
> ---
> tools/perf/builtin-daemon.c | 9 +++++----
> 1 file changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/tools/perf/builtin-daemon.c b/tools/perf/builtin-daemon.c
> index 83954af36753..0b6ffd51c475 100644
> --- a/tools/perf/builtin-daemon.c
> +++ b/tools/perf/builtin-daemon.c
> @@ -23,6 +23,7 @@
> #include <sys/signalfd.h>
> #include <sys/wait.h>
> #include <poll.h>
> +#include <inttypes.h>
> #include "builtin.h"
> #include "perf.h"
> #include "debug.h"
> @@ -688,7 +689,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
> /* lock */
> csv_sep, daemon->base, "lock");
>
> - fprintf(out, "%c%lu",
> + fprintf(out, "%c%" PRIu64 "",
> /* session up time */
> csv_sep, (curr - daemon->start) / 60);
>
> @@ -700,7 +701,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
> daemon->base, SESSION_OUTPUT);
> fprintf(out, " lock: %s/lock\n",
> daemon->base);
> - fprintf(out, " up: %lu minutes\n",
> + fprintf(out, " up: %" PRIu64 " minutes\n",
> (curr - daemon->start) / 60);
> }
> }
> @@ -727,7 +728,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
> /* session ack */
> csv_sep, session->base, SESSION_ACK);
>
> - fprintf(out, "%c%lu",
> + fprintf(out, "%c%" PRIu64 "",
> /* session up time */
> csv_sep, (curr - session->start) / 60);
>
> @@ -745,7 +746,7 @@ static int cmd_session_list(struct daemon *daemon, union cmd *cmd, FILE *out)
> session->base, SESSION_CONTROL);
> fprintf(out, " ack: %s/%s\n",
> session->base, SESSION_ACK);
> - fprintf(out, " up: %lu minutes\n",
> + fprintf(out, " up: %" PRIu64 " minutes\n",
> (curr - session->start) / 60);
> }
> }

Sorry I missed this earlier, but IIUC this one is actually incorrect: on
most 32-bit platforms time_t is 32 bits, it was later extended to 64
bits. RISC-V is special because the work to make time_t 64-bit had
started when we submitted the port, so we just jumped straight to the
legacy-free uABI (after some headaches).

So IIUC this would introduce a warning for some other targets. Either
way I shouldn't have picked it up as it's not a RISC-V patch (this is a
generic perf file), so I'm going to drop it from fixes.

Sorry for the confusion!