On Thu, Dec 15, 2016 at 7:53 AM, Branko Čibej <[email protected]> wrote:
> Some warnings on trunk have been getting on my nerves and I finally got
> to the point of starting to fix them. A couple are not entirely clear-cut:
>
> 1.
>
> buckets/hpack_buckets.c:1872:27: warning: comparison of constant
> 18446744073709551615 with
> expression of type 'apr_uint32_t' (aka 'unsigned int') is always
> false [-Wtautological-constant-out-of-range-compare]
> if (v >= APR_SIZE_MAX)
> ~ ^ ~~~~~~~~~~~~
>
>
> The relevant code is:
>
> {
> apr_uint32_t v;
>
> status = read_hpack_int(&v, NULL, bucket, 5);
> if (status)
> continue;
>
> /* Send remote tablesize update to our table */
> if (v >= APR_SIZE_MAX)
> return SERF_ERROR_HTTP2_COMPRESSION_ERROR;
> status = hpack_table_size_update(ctx->tbl,
> (apr_size_t)v);
> if (status)
> return status;
>
> ctx->state = HPACK_DECODE_STATE_INITIAL;
> break;
> }
>
>
> As far as I can see, not only is the condition never true, but
> read_hpack_int() already checks for overflow and returns the same status
> code if it detects it. I'd suggest just removing this check, but maybe
> I'm missing something.
>
> 2.
>
> buckets/fcgi_buckets.c:746:48: warning: format specifies type 'unsigned
> int' but the argument
> has type 'apr_size_t' (aka 'unsigned long') [-Wformat]
> ctx->frame_type, ctx->stream_id, payload);
> ^~~~~~~
>
>
> The relevant code is:
>
> serf__log(LOGLVL_DEBUG, LOGCOMP_CONN, __FILE__, ctx->config,
> "Generating 0x%x frame on stream 0x%x of size 0x%x\n",
> ctx->frame_type, ctx->stream_id, payload);
>
> I know APR doesn't provide a hex variant of APR_SIZE_T_FMT; on the other
> hand, writing the /size/ of something to the log in hex is hardly
> friendly to the poor lass who ends up having to read it. I propose
> changing that size to decimal and using APR_SIZE_T_FMT. The alternative
> is to use APR_UINT64_T_HEX_FMT and adding a cast.
>
For the payload variable, yeah, that's supposed to be
logged
as a
decimal
number instead of a hex
, so agreed with your proposal.
>
> -- Brane
>
Lieven
