Changeset 2090 for trunk/src/emx/ChangeLog.LIBC

Timestamp:

Jun 27, 2005, 5:30:23 AM (20 years ago)

Author:

bird

Message:

o Rewrote the libcXXXX.dll install rule to not play with emxload since

emxload sticks around when first loaded.

o Made unsetenv() cause tzset reload just like putenv() and setenv().
o Changed putenv() behaviour for removal (no '=' in input) to be identical

to unsetenv(). This solves duplicate variable testcase.

o Changed filestreams opened for both reading and writing to switch

more willingly between read and write mode.

o Made sscan() return EOF on EOF.
o Made sscan() skip blanks in input for '%%'.
o Implemented the stdio_unlocked stuff with BSD and GLIBC extension.

TODO: Require a new locktype which support recursive usage! Currently

a hack using 8 free flag bits is used.

o Corrected realpath() behaviour to return the path up to the failure

point on error. _realrealpath() also does this now.
(This also saves us a temp buffer and a copy of the result.)

o Fixed a bug in the handling of ".." in the path resolver.
o Extended that path resolver to check the directoryness of the input

according to flags and any trailing slashes.

o Change the path resolver to be very strict about UNC - UNC have exactly

two slashes at the start of the path.

File:

• trunk/src/emx/ChangeLog.LIBC (modified) (1 diff, 1 prop)

trunk/src/emx/ChangeLog.LIBC

@@ -1 +1 @@
 /* $Id$ */
+
+2005-06-26: knut st. osmundsen <bird-gccos2-spam@anduin.net>
+    - libc:
+        o Added [__]isinf[fl] which seems to have been mislaid during the
+          math source import.
+        o Rewrote the libcXXXX.dll install rule to not play with emxload since
+          emxload sticks around when first loaded.
+        o Made unsetenv() cause tzset reload just like putenv() and setenv().
+        o Changed putenv() behaviour for removal (no '=' in input) to be identical
+          to unsetenv(). This solves duplicate variable testcase.
+        o Changed filestreams opened for both reading and writing to switch
+          more willingly between read and write mode.
+        o Made sscan() return EOF on EOF.
+        o Made sscan() skip blanks in input for '%%'.
+        o Implemented the stdio_unlocked stuff with BSD and GLIBC extension.
+          TODO: Require a new locktype which support recursive usage! Currently
+                a hack using 8 free flag bits is used.
+        o Corrected realpath() behaviour to return the path up to the failure
+          point on error. _realrealpath() also does this now.
+          (This also saves us a temp buffer and a copy of the result.)
+        o Fixed a bug in the handling of ".." in the path resolver.
+        o Extended that path resolver to check the directoryness of the input
+          according to flags and any trailing slashes.
+        o Change the path resolver to be very strict about UNC - UNC have exactly
+          two slashes at the start of the path.
 
 2005-06-25: knut st. osmundsen <bird-gccos2-spam@anduin.net>