[pacman-dev] [PATCH] libalpm: add PID to db.lck

Allan McRae allan at archlinux.org
Mon Nov 3 09:01:24 EST 2008


Dan McGee wrote:
> On Sun, Nov 2, 2008 at 8:22 PM, Allan McRae <allan at archlinux.org> wrote:
>   
>> This is the first step in being able to automatically remove phantom
>> lock files.
>>
>> Signed-off-by: Allan McRae <allan at archlinux.org>
>> ---
>>  lib/libalpm/util.c |   15 +++++++++++++--
>>  1 files changed, 13 insertions(+), 2 deletions(-)
>>
>> diff --git a/lib/libalpm/util.c b/lib/libalpm/util.c
>> index da3463b..811572a 100644
>> --- a/lib/libalpm/util.c
>> +++ b/lib/libalpm/util.c
>> @@ -242,7 +242,9 @@ char *_alpm_strreplace(const char *str, const char *needle, const char *replace)
>>  int _alpm_lckmk()
>>  {
>>        int fd;
>> -       char *dir, *ptr;
>> +       pid_t pid;
>> +       size_t len;
>> +       char *dir, *ptr, *spid;
>>        const char *file = alpm_option_get_lockfile();
>>
>>        /* create the dir of the lockfile first */
>> @@ -256,7 +258,16 @@ int _alpm_lckmk()
>>
>>        while((fd = open(file, O_WRONLY | O_CREAT | O_EXCL, 0000)) == -1
>>                        && errno == EINTR);
>> -       return(fd > 0 ? fd : -1);
>> +       if(fd > 0) {
>> +               pid = getpid();
>> +               size_t len = snprintf(spid, 0, "%d", pid) + 1;
>>     
> I think we need to use %ld here as a pid_t can be a signed type, up to
> length long. Thus the whole thing:
> snprintf(spid, 0, "%ld", (long)pid)
> Obviously do this below too. And do you want to add a newline char in
> here to make your life easier when reading this in?
>
>   
>> +               spid = malloc(len);
>> +               snprintf(spid, len, "%d", pid) + 1;
>>     
> Think you left a +1 here by accident.
>
>   
>> +               while(write(fd, (void *)spid, len) == -1 && errno == EINTR);
>> +               fsync(fd);
>> +               return(fd);
>> +       }
>> +       return(-1);
>>  }
>>     
>
> And re: your second message, I think your write() call is just fine here.

There is an updated version with all these comments fixed on my working 
branch 
(http://dev.archlinux.org/~allan/gitweb/gitweb.cgi?p=pacman.git;a=commitdiff;h=8f6e71fe).  
As a bonus, I also fixed the memory leak...

Allan






More information about the pacman-dev mailing list