This is the mail archive of the
mailing list for the GCC project.
Re: 3.2 PATCH: Ada parallel bootstrap fixes
- From: Florian Weimer <fw at deneb dot enyo dot de>
- To: Geert Bosch <bosch at gnat dot com>
- Cc: Zack Weinberg <zack at codesourcery dot com>, Robert Dewar <dewar at gnat dot com>, kenner at vlsi1 dot ultra dot nyu dot edu, pkoning at equallogic dot com, gcc at gcc dot gnu dot org
- Date: Sun, 19 May 2002 21:09:07 +0200
- Subject: Re: 3.2 PATCH: Ada parallel bootstrap fixes
- References: <049B71BE-6B2A-11D6-AF14-00039344BF4A@gnat.com>
Geert Bosch <email@example.com> writes:
> On Saturday, May 18, 2002, at 01:56 , Zack Weinberg wrote:
>> In this particular case, I think the best thing to do is copy the
>> mktemp() code out of GNU libc into the Ada runtime. That code is
I think you mean "mkstemp()".
>> known to work, and we don't have to worry about the system library not
>> providing mktemp or its being buggy.
> Does this allow retrieval of the filename?
Yes, it does. After all, you have to supply a name if you want to
create a file using the kernel interfaces.
> My reading of the Ada standard requires one to be able to retrieve
> the filename if there is a name through which the file can be
We can create the file with no-access permissions and unlink it just
after that, then it is never accessible by a name, but ...
> As Robert confirmed, there is also code out there depending on being
> able to retrieve the filename.
... Robert says we can't do that, so we
> If you can come up with a patch that does fix the problem and allows
> the user to obtain the external filename, that would be great.
I'm going to submit a patch soon which moves max_path_len from
cstream.c to __gnat_max_path_len in adaint.c, so that we can remove
the buffer overflow in a clean way first.