clvmd.c: Remove redundant if-before-free tests.
PATCH 1/1 clvmd.c: Remove redundant if-before-free tests.
Since "free (NULL);" is universally safe these days, I propose to remove all of lvm's redundant "if-before-free" tests. There are only 5. If no one objects, I'll commit it tonight.
In case you need more justification than "POSIX requires it", (understandable, that :-), the following was enough for Wine development a year and a half ago:
http://www.winehq.org/pipermail/wine-patches/2006-October/031544.html
Sure, the extra test is not a big deal, but removing it does make the code slightly smaller, and maybe even more readable (people won't wonder about the test of such a pointer, when they know that free can handle a NULL pointer).
The changes in the patch below were created by running this command after a fresh "clone":
find . -name '*.c' -print0 | xargs -0 perl -0x3b -pi -e \
's/\bif\s*\(\s*(\S+?)(?:\s*!=\s*NULL)?\s*\)\s+(free\s*\(\s*\1\s*\))/$2/s'daemons/clvmd/clvmd.c +5 -10
@@ -903,8 +903,7 @@ static int read_from_local_sock(struct local_client *thisfd) }/* Free the command buffer */ -if (thisfd->bits.localsock.cmd) -free (thisfd->bits.localsock.cmd); +free (thisfd->bits.localsock.cmd);/* Clear out the cross-link */ if (thisfd->bits.localsock.pipe_client !=NULL )
@@ -939,8 +938,7 @@ static int read_from_local_sock(struct local_client *thisfd) }/* Free any old buffer space */ -if (thisfd->bits.localsock.cmd) -free (thisfd->bits.localsock.cmd); +free (thisfd->bits.localsock.cmd);/* See if we have the whole message */ argslen =
@@ -1547,8 +1545,7 @@ static void send_local_reply(struct local_client *client, int status, int fd) } thisreply = thisreply->next; -if (tempreply->replymsg) -free (tempreply->replymsg); +free (tempreply->replymsg);free (tempreply); }
@@ -1579,8 +1576,7 @@ static void free_reply(struct local_client *client) thisreply = thisreply->next; -if (tempreply->replymsg) -free (tempreply->replymsg); +free (tempreply->replymsg);free (tempreply); } client->bits.localsock.replies =NULL ;
@@ -1732,8 +1728,7 @@ static __attribute__ ((noreturn)) void *lvm_thread_fn(void *arg)pthread_mutex_unlock (&lvm_thread_mutex);process_work_item (cmd); -if (cmd->msg) -free (cmd->msg); +free (cmd->msg);free (cmd);pthread_mutex_lock (&lvm_thread_mutex);
--
1.5.4.1.98.gf3293
PATCH 1/1 Remove redundant if-before-free tests.
[as I just wrote to lvm-devel:
http://www.redhat.com/archives/lvm-devel/2008-February/msg00017.html ]
Quoted 8 lines · Jim Meyering
Since "free (NULL);" is universally safe these days, I propose to remove all of lvm's redundant "if-before-free" tests. If no one objects, I'll commit it tonight.
In case you need more justification than "POSIX requires it", (understandable, that :-), the following was enough for Wine development a year and a half ago:
Quoted 6 lines
http://www.winehq.org/pipermail/wine-patches/2006-October/031544.htmlSure, the extra test is not a big deal, but removing it does make the code slightly smaller, and maybe even more readable (people won't wonder about the test of such a pointer, when they know that free can handle a NULL pointer).
Same for dm, but here there are only 3:
--- dmeventd/old/dmevent.c | 6 ++---- multilog/pthread_lock/pthread_lock.c | 3 +-- 2 files changed, 3 insertions(+), 6 deletions(-)
dmeventd/old/dmevent.c +2 -4
@@ -220,10 +220,8 @@ int main(int argc, char **argv) out:multilog_del_type (standard); -if (device) -free (device); -if (dso_name) -free (dso_name); +free (device); +free (dso_name);exit (ret); }
multilog/pthread_lock/pthread_lock.c +1 -2
@@ -62,7 +62,6 @@ int unlock_fn(void *handle)void destroy_locking (void *handle) {pthread_mutex_t *mutex = (pthread_mutex_t *)handle; -if (mutex) -free (mutex); +free (mutex); }
--
1.5.4.1.98.gf3293
Alasdair G Kergon
maintainerRe: [lvm-devel] [PATCH] Remove redundant if-before-free tests.
Quoted 1 line · Jim Meyering
Same for dm, but here there are only 3:
2 of which are in obsolete code - don't change those (the dir will get dropped when we combine the trees); the other of which is in unfinished & unreviewed code (yes, change that one)
Alasdair
Signature · 1 line
[email protected]