Paolo Bonzini <[EMAIL PROTECTED]> writes:
>> I do feel that the entire unlocked IO feature is poorly implemented in
>> the way that it requires changes to all code that wants to use it, but I
>> don't have a solution for that. Can't you set some global flag that
>> makes all getc calls unlocked?
- Original Message -
From: "Eric Blake" <[EMAIL PROTECTED]>
To:
Sent: Wednesday, August 27, 2008 10:38 AM
Subject: Re: GNUmakefile: git-version-gen, maintainer-check, install
> Denny Joel E cs.clemson.edu> writes:
>
>> > Well, I tried to make it a hard error, so that it would be the
> I do feel that the entire unlocked IO feature is poorly implemented in
> the way that it requires changes to all code that wants to use it, but I
> don't have a solution for that. Can't you set some global flag that
> makes all getc calls unlocked? Instead of modifying all code to call
> getc_u
Paolo Bonzini <[EMAIL PROTECTED]> writes:
>> However, to deal with 2) and 3) I would prefer if the current code stays
>> around. It could be moved to a 'getline-portable' module.
>
> For now, what do you think about the other patch? It does already
> provide an interesting performance increase.
Denny Joel E cs.clemson.edu> writes:
> > Well, I tried to make it a hard error, so that it would be the only
> > output, but others on the list convinced me to reduce it back to a
> > warning. I don't know of any GNU make construct to make it appear last.
> > Maybe we just insert a sleep 10 afte
> However, to deal with 2) and 3) I would prefer if the current code stays
> around. It could be moved to a 'getline-portable' module.
For now, what do you think about the other patch? It does already
provide an interesting performance increase.
Paolo
Paolo Bonzini <[EMAIL PROTECTED]> writes:
> The number of deletions makes it yummy...
I like it, although in order to deprecate the current implementation,
I'd like to see getndelim2 solve this problem:
1) LGPL license compatibility
2) Portability: getndelim2 uses freadptr and freedseek which a
Eric Blake wrote:
> According to Paolo Bonzini on 8/27/2008 6:54 AM:
>> Except that getndelim2 and memchr2 are GPL, while freadptr and freadseek
>> are LGPL (I assume v3+). Would it be okay to change all four to LGPLv2+?
>
> I'm okay with memchr2 being LGPLv2+; it does little more than memchr,
>
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Paolo Bonzini on 8/27/2008 6:54 AM:
>
> Except that getndelim2 and memchr2 are GPL, while freadptr and freadseek
> are LGPL (I assume v3+). Would it be okay to change all four to LGPLv2+?
I'm okay with memchr2 being LGPLv2+; it does lit
>> Sounds interesting. How does this compare with the current approach used
>> by getndelim2, which uses freadptr to avoid getc?
>
> I guess there's also this other possibility.
>
> lib/getdelim.c | 107 ++-
> m4/getdelim.m4 | 10 -
> modu
Eric Blake wrote:
> According to Paolo Bonzini on 8/27/2008 5:48 AM:
>> Under MacOS, getc is not a macro, only getc_unlocked is. This patch
>> does two things: 1) it makes getdelim use getc_unlocked if it can wrap
>> the calls with flockfile/funlockfile; 2) it makes getdelim omit the
>> locking al
Eric Blake wrote:
> According to Paolo Bonzini on 8/27/2008 5:48 AM:
>> Under MacOS, getc is not a macro, only getc_unlocked is. This patch
>> does two things: 1) it makes getdelim use getc_unlocked if it can wrap
>> the calls with flockfile/funlockfile; 2) it makes getdelim omit the
>> locking al
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Paolo Bonzini on 8/27/2008 5:48 AM:
> Under MacOS, getc is not a macro, only getc_unlocked is. This patch
> does two things: 1) it makes getdelim use getc_unlocked if it can wrap
> the calls with flockfile/funlockfile; 2) it makes getdeli
Under MacOS, getc is not a macro, only getc_unlocked is. This patch
does two things: 1) it makes getdelim use getc_unlocked if it can wrap
the calls with flockfile/funlockfile; 2) it makes getdelim omit the
locking altogether if unlocked-io is in effect. It speeds up sed by
almost 2x in very simp
14 matches
Mail list logo
