[PATCH] compat/regex/regexec.c: Fix some sparse warnings

Ramsay Jones· May 17, 2011, 17:40 UTC · lore

In particular, sparse issues the following warnings:

        SP compat/regex/regex.c
    .../regexec.c:2329:40: warning: Using plain integer as NULL pointer
    .../regexec.c:2316:15: warning: symbol 'merge_state_with_log' was \
        not declared. Should it be static?
    .../regexec.c:2395:15: warning: symbol 'find_recover_state' was not \
        declared. Should it be static?

In order to fix the "symbol not declared" warnings (they are in fact declared static earlier in the file), we simply add the static modifier to the function definitions.

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
---

Hi Junio,

Now that I'm building git with NO_REGEX=1 in my config.mak, I'm also seeing these sparse warnings.

Note that the "symbol not declared" warnings actually highlight a sparse bug; the earlier static function declarations should have (effectively) 'transferred' the static modifier to the corresponding function definition ... [I have a testcase for this in my sparse repo, but I don't seem to have found time to actually create a patch to fix this. :(]

ATB,
Ramsay Jones
 
 compat/regex/regexec.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

Show changes to compat/regex/regexec.c +3 −3

diff --git a/compat/regex/regexec.c b/compat/regex/regexec.c
index 0194965..0cd6e0e 100644
--- a/compat/regex/regexec.c
+++ b/compat/regex/regexec.c
@@ -2313,7 +2313,7 @@ transit_state (reg_errcode_t *err, re_match_context_t *mctx,
 }
 
 /* Update the state_log if we need */
-re_dfastate_t *
+static re_dfastate_t *
 internal_function
 merge_state_with_log (reg_errcode_t *err, re_match_context_t *mctx,
 		      re_dfastate_t *next_state)
@@ -2326,7 +2326,7 @@ merge_state_with_log (reg_errcode_t *err, re_match_context_t *mctx,
       mctx->state_log[cur_idx] = next_state;
       mctx->state_log_top = cur_idx;
     }
-  else if (mctx->state_log[cur_idx] == 0)
+  else if (mctx->state_log[cur_idx] == NULL)
     {
       mctx->state_log[cur_idx] = next_state;
     }
@@ -2392,7 +2392,7 @@ merge_state_with_log (reg_errcode_t *err, re_match_context_t *mctx,
 /* Skip bytes in the input that correspond to part of a
    multi-byte match, then look in the log for a state
    from which to restart matching.  */
-re_dfastate_t *
+static re_dfastate_t *
 internal_function
 find_recover_state (reg_errcode_t *err, re_match_context_t *mctx)
 {

-- 
1.7.5

Ævar Arnfjörð Bjarmason· May 17, 2011, 22:52 UTC · re: Ramsay Jones · lore

Re: [PATCH] compat/regex/regexec.c: Fix some sparse warnings

On Tue, May 17, 2011 at 19:40, Ramsay Jones <ramsay@ramsay1.demon.co.uk> wrote:

> In particular, sparse issues the following warnings:

Aside from this specific change, I think it's a problem that we pull in external libraries, then apply patches to them that upstream would probably want, don't tell upstream, and will inevitably have to deal with conflict when we merge in new versions from upstream.

Maybe we should have some file with contact details for these compat libraries, and try to resolve issues with upstream if we need to apply patches like this one?

Junio C Hamano· May 18, 2011, 04:04 UTC · re: Ramsay Jones · lore

Re: [PATCH] compat/regex/regexec.c: Fix some sparse warnings

Ramsay Jones <ramsay@ramsay1.demon.co.uk> writes:

> In particular, sparse issues the following warnings:
>
>         SP compat/regex/regex.c

I'd rather leave stuff in compat/ that were imported from other people intact, unless they are functionally wrong. We don't fix their styles either.

Replacing our copy with newer versions that have fixed these issues from upstreams is Ok, though.

patchcompat/regex/regexec.c: Fix some sparse warnings