Re: [Qemu-devel] [PATCH v3 RESEND 3/3] nvram: at24c: use standard error reporting

[Peter Maydell <peter.maydell@linaro.org>]

On 20 May 2018 at 08:00, Wolfram Sang <wsa+renesas@sang-engineering.com> wrote:
> Replace the ERR macro with error_report() because fprintf is deprecated.
> This also fixes the prefix printed out twice.
>
> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
> ---
>  hw/nvram/eeprom_at24c.c | 17 ++++++-----------
>  1 file changed, 6 insertions(+), 11 deletions(-)
> @@ -63,8 +61,7 @@ int at24c_eeprom_event(I2CSlave *s, enum i2c_event event)
>          if (ee->blk && ee->changed) {
>              int len = blk_pwrite(ee->blk, 0, ee->mem, ee->rsize, 0);
>              if (len != ee->rsize) {
> -                ERR(TYPE_AT24C_EE
> -                        " : failed to write backing file\n");
> +                error_report("failed to write backing file");
>              }
>              DPRINTK("Wrote to backing file\n");
>          }
> @@ -125,7 +122,7 @@ int at24c_eeprom_init(I2CSlave *i2c)
>      EEPROMState *ee = AT24C_EE(i2c);
>
>      if (!ee->rsize) {
> -        ERR("rom-size not allowed to be 0\n");
> +        error_report("rom-size not allowed to be 0");
>          exit(1);
>      }

Hi; if we're going to overhaul the error handling for this
device, I think we might as well do it properly, by moving
the code in this init function which can fail into a new
device realize method. The realize method takes an Error**
and can report failures that way rather than by causing
QEMU to exit.

> @@ -166,8 +162,7 @@ void at24c_eeprom_reset(DeviceState *state)
>          int len = blk_pread(ee->blk, 0, ee->mem, ee->rsize);
>
>          if (len != ee->rsize) {
> -            ERR(TYPE_AT24C_EE
> -                    " : Failed initial sync with backing file\n");
> +            error_report("Failed initial sync with backing file");
>          }
>          DPRINTK("Reset read backing file\n");
>      }

Errors in reset and event methods are a bit more awkward;
error_report is an improvement in those cases.

thanks
-- PMM