Re: [PATCH v2 3/3] mingw: replace isatty() hack
- From
Johannes Sixt <j6t@kdbg.org>
- Date
- Dec 22, 2016, 22:28 UTC
- Message-ID
- <04859cf9-e67a-28ab-ccb3-249687e696c8@kdbg.org>
- In-Reply-To
- <alpine.DEB.2.20.1612222235200.155951@virtualbox>
Am 22.12.2016 um 22:37 schrieb Johannes Schindelin:
Show 56 quoted lines
> Hi Hannes,
>
> On Thu, 22 Dec 2016, Johannes Sixt wrote:
>
>> Am 22.12.2016 um 18:09 schrieb Johannes Schindelin:
>>> +static HANDLE swap_osfhnd(int fd, HANDLE new_handle)
>>> +{
>>> + /*
>>> + * Create a copy of the original handle associated with fd
>>> + * because the original will get closed when we dup2().
>>> + */
>>> + HANDLE handle = (HANDLE)_get_osfhandle(fd);
>>> + HANDLE duplicate = duplicate_handle(handle);
>>>
>>> + /* Create a temp fd associated with the already open "new_handle". */
>>> + int new_fd = _open_osfhandle((intptr_t)new_handle, O_BINARY);
>>>
>>> + assert((fd == 1) || (fd == 2));
>>>
>>> + /*
>>> + * Use stock dup2() to re-bind fd to the new handle. Note that
>>> + * this will implicitly close(1) and close both fd=1 and the
>>> + * originally associated handle. It will open a new fd=1 and
>>> + * call DuplicateHandle() on the handle associated with new_fd.
>>> + * It is because of this implicit close() that we created the
>>> + * copy of the original.
>>> + *
>>> + * Note that the OS can recycle HANDLE (numbers) just like it
>>> + * recycles fd (numbers), so we must update the cached value
>>> + * of "console". You can use GetFileType() to see that
>>> + * handle and _get_osfhandle(fd) may have the same number
>>> + * value, but they refer to different actual files now.
>>
>> Certainly, the OS does not recycle handle values that are in use (open). Then
>> I do not quite get the point of this paragraph. See...
>>
>>> + *
>>> + * Note that dup2() when given target := {0,1,2} will also
>>> + * call SetStdHandle(), so we don't need to worry about that.
>>> + */
>>> + dup2(new_fd, fd);
>>> + if (console == handle)
>>> + console = duplicate;
>>
>> ... This is where "the cached value of console is updated", right? If console
>> == handle, then this is not because a handle value was recycled, but because
>> fd *was* console. Since the old value of console has been closed by the
>> dup2(), we must refer to the back-up value in the future. Am I missing
>> something?
>
> You are correct, we must update `console` because `handle` is no longer
> the handle we want.
>
> The comment above only meant to reinforce that we have to forget about the
> previous handle, too, as we might access something completely different
> than a console otherwise.It is like accessing a pointer after free().
When I read the paragraph for the first time, I expected some information to be saved, then some handles to be closed and re-opened, which would possibly recycle a handle value, and then same state to be resurrected depending on the information saved earlier.
But nothing of this kind happens, only:
dup2(new_fd, fd); if (console == handle) console = duplicate;
which is necessary and, once one has understood the context, obvious.
> Would you have a suggestion how to rephrase the comment to make it less > confusing?
Perhaps
* This might close the cached console handle. * We must cache the live duplicate instead.
Then update the cache before the dup2, because at this time all 3 of console, handle, and duplicate are live and cannot be recycled:
if (console == handle) console = duplicate; dup2(new_fd, fd);
-- Hannes