Re: Format string bugs & race conditions

[Kilgallen@SpamCop.net (Larry Kilgallen)]

In article <mailman.13.1098016133.10401.comp.lang.ada@ada-france.org>, Marius Amado Alves <amado.alves@netcabo.pt> writes:
>> /*  If lockfile already exists, wait until it goes away.  */
>> while((p = fopen("lockfile", "r"))) {
>>     fclose(p);
>>     sleep(1);
>> }
>> /*  (race)  */
>> p = fopen("lockfile", "w");
>> /*  Access the stuff which lockfile is supposed to protect from
>>  *  multiple access.  But of course it doesn't.  */
>> fclose(p);
>> remove("lockfile");
>> 
>> Ada just makes it easier to implement this bug within a single
>> instance of the executable.
> 
> No. Using files for locks is equally easy to implement and equally buggy 
> in any language.

And that might be "not buggy at all", depending on _how_ you use files.

Regardless of programming language*, when one calls the VMS system
service SYS$CREATE with the FAB$V_CIF option a status code of
RMS$_CREATED indicates that the file was created by this process.
If not, it is time to try again for the example above.  Of course
opening a file is a costly synchronization mechanism which should
only be used in cases where the cost will be dwarfed by the cost
of the operation being protected.

      *	As usual Ada programmers use slightly different symbol names
	since Ada does not allow "$" in symbols.