Do not use osutil.c with python 2.4 and Windows (
issue1364)
Windows python 2.4 os.stat() reports times including DST offset, while osutil.c
reports the correct value, which makes status() systematically compare files
content. This bug is fixed in python 2.5. Using osutil.py instead of osutil.c
is 4x times slower on large repositories but current code is completely
unusable. Given few people are likely to use python 2.4 on Windows this
solution was considered a good trade-off compared to more invasive solutions
trying to address the offset issue.
adding a
% default context
[0;1mdiff -r cf9f4ba66af2 a[0m
[0;31;1m--- a/a[0m
[0;32;1m+++ b/a[0m
[0;35m@@ -2,7 +2,7 @@[0m
c
a
a
[0;31m-b[0m
[0;32m+dd[0m
a
a
c
% --unified=2
[0;1mdiff -r cf9f4ba66af2 a[0m
[0;31;1m--- a/a[0m
[0;32;1m+++ b/a[0m
[0;35m@@ -3,5 +3,5 @@[0m
a
a
[0;31m-b[0m
[0;32m+dd[0m
a
a
% diffstat
a | 2 [0;32m+[0m[0;31m-[0m
1 files changed, 1 insertions(+), 1 deletions(-)
% record
[0;1mdiff --git a/a b/a[0m
[0;36;1mold mode 100644[0m
[0;36;1mnew mode 100755[0m
1 hunks, 2 lines changed
examine changes to 'a'? [Ynsfdaq?] [0;35m@@ -2,7 +2,7 @@[0m
c
a
a
[0;31m-b[0m
[0;32m+dd[0m
a
a
c
record this change to 'a'? [Ynsfdaq?]
rolling back last transaction
% qrecord
[0;1mdiff --git a/a b/a[0m
[0;36;1mold mode 100644[0m
[0;36;1mnew mode 100755[0m
1 hunks, 2 lines changed
examine changes to 'a'? [Ynsfdaq?] [0;35m@@ -2,7 +2,7 @@[0m
c
a
a
[0;31m-b[0m
[0;32m+dd[0m
a
a
c
record this change to 'a'? [Ynsfdaq?]