Re: [PATCH 2/3] Introduce scoped_mmapped_file

[Simon Marchi <simark@simark.ca>]

On 2018-07-23 10:17 PM, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi <simon.marchi@ericsson.com> writes:
> 
> Simon> We already have scoped_mmap, which can is a thin RAII layer over mmap.
> Simon> If one simply wants to mmap an entire file for reading, it takes a bit
> Simon> of boilerplate.  This patch introduces the scoped_mmapped_file class to
> Simon> make this easier.
> 
> Simon> +  /* Map FILENAME in memory.  Throw an error if anything goes wrong.  */
> Simon> +  scoped_mmapped_file (const char *filename)
> Simon> +  {
> Simon> +    m_fd.reset (open (filename, O_RDONLY));
> Simon> +    if (m_fd.get () < 0)
> Simon> +      perror_with_name ("open");
> Simon> +
> Simon> +    off_t size = lseek (m_fd.get (), 0, SEEK_END);
> Simon> +    if (size < 0)
> Simon> +      perror_with_name ("lseek");
> Simon> +
> Simon> +    /* We can't map an empty file.  */
> Simon> +    if (size == 0)
> Simon> +      error (_("file to mmap is empty"));
> Simon> +
> Simon> +    m_mmap.reset (nullptr, size, PROT_READ, MAP_PRIVATE, m_fd.get (), 0);
> Simon> +    if (m_mmap.get () == MAP_FAILED)
> Simon> +      perror_with_name ("mmap");
> 
> It seems to me that there's no need to keep the fd open after the mmap,
> so this could be done by just having the scoped_fd as a local variable,
> and then I suppose the scoped_fd changes wouldn't be needed either.
> 
> Tom
> 

Ah, I didn't know the fd didn't need to be kept open, but you're right:
https://stackoverflow.com/questions/17490033/do-i-need-to-keep-a-file-open-after-calling-mmap-on-it

I'll try that.

Simon