back for now

back for now

[Ulrich Drepper]

I'm back for now and checked in the changes I've made over the days.
Mainly I've taken most of the changes I got and applied them (some of
them modified, e.g., HJ's thread patch).

Besides I've reworked dcgettext (the caching is now done in the
function itself, not in the wrapper macro).  This saves lots of bss
and code space in applications and it also enables gcc to perform the
-Wformat tests.  The changes are not tested much but at least the test
suite succeeds.

I've also adjusted the libm ULPs.  I'm not sure whether it was the
Mobile PII I have in my laptop or the gcc version (I think the
former).  Anyway, the results for some functions are worse.

Last but not least, I've finished writing the LC_COLLATE parsing in
localedef, corrected some bugs, and started writing the code to write
the data to file.  We are not too far away from the complete locale
implementation.


I mentioned last weekend that I would be able to handle patches during
the week but it turned out that the network connections from this
"conference" were really bad.  They set up a network and had no free
DHCP addresses.  Clever.  Even more because the attendance number was
so low that at most 20% of the machines in the tutorials were used.
There were probably more speakers at the conference than paying
attendees.  I cannot blame them, NYC is far too expensive.


Today I'll work on all the patches for 2.1 and not processed patches
for 2.2.

-- 
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------

Re: back for now

[H . J . Lu]

On Sat, Dec 18, 1999 at 11:53:25AM -0800, Ulrich Drepper wrote:
> I'm back for now and checked in the changes I've made over the days.
> Mainly I've taken most of the changes I got and applied them (some of
> them modified, e.g., HJ's thread patch).
> 

Can we also fix linuxthread in 2.1? Also I am not quite sure if we should
put a guard page beyond the stack bottom. Shouldn't we use this or at least
fix the comment?



H.J.
--- manager.c	Sun Dec 19 07:51:27 1999
+++ /tmp/manager.c	Sun Dec 19 08:17:30 1999
@@ -315,7 +315,7 @@ static int pthread_allocate_stack(const 
       else
         {
           /* Put a bad page at the bottom of the stack */
-          guardaddr = (void *)new_thread_bottom - stacksize;
+          guardaddr = (void *)new_thread_bottom;
           guardsize = attr->__guardsize;
           if (mmap ((caddr_t) guardaddr, guardsize, 0, MAP_FIXED, -1, 0)
               == MAP_FAILED)

A guard patch for linuxthreads

[H . J . Lu]

On Sun, Dec 19, 1999 at 08:21:16AM -0800, H . J . Lu wrote:
> On Sat, Dec 18, 1999 at 11:53:25AM -0800, Ulrich Drepper wrote:
> > I'm back for now and checked in the changes I've made over the days.
> > Mainly I've taken most of the changes I got and applied them (some of
> > them modified, e.g., HJ's thread patch).
> > 
> 
> Can we also fix linuxthread in 2.1? Also I am not quite sure if we should
> put a guard page beyond the stack bottom. Shouldn't we use this or at least
> fix the comment?
> 
> 

I believe this patch is what we want.


H.J.
--- manager.c	Sun Dec 19 07:51:27 1999
+++ /tmp/manager.c	Sun Dec 19 08:43:59 1999
@@ -315,8 +315,8 @@ static int pthread_allocate_stack(const 
       else
         {
           /* Put a bad page at the bottom of the stack */
-          guardaddr = (void *)new_thread_bottom - stacksize;
           guardsize = attr->__guardsize;
+          guardaddr = (void *)new_thread_bottom - guardsize;
           if (mmap ((caddr_t) guardaddr, guardsize, 0, MAP_FIXED, -1, 0)
               == MAP_FAILED)
             {

More on linuxthreads

[H . J . Lu]

On Sat, Dec 18, 1999 at 11:53:25AM -0800, Ulrich Drepper wrote:
> I'm back for now and checked in the changes I've made over the days.
> Mainly I've taken most of the changes I got and applied them (some of
> them modified, e.g., HJ's thread patch).
> 

There are new 64 bit versions of I/O functions in glibc 2.1. But there
are nothing for them in linuxthreads.


H.J.

Re: More on linuxthreads

[Ulrich Drepper]

"H . J . Lu" <hjl@lucon.org> writes:

> There are new 64 bit versions of I/O functions in glibc 2.1. But there
> are nothing for them in linuxthreads.

I have no idea what you mean.  wrapsyscall.c defines all the 64
versions as well.

-- 
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------