Thread (1 message) 1 message, 1 author, 2018-10-24

Re: [PATCH] test: avoid failures when USE_NED_ALLOCATOR

From: Junio C Hamano <hidden>
Date: 2018-10-24 03:10:38

Carlo Marcelo Arenas Belón  [off-list ref] writes:
contrib/nedmalloc doesn't support MALLOC_CHECK_ or MALLOC_PERTURB_
so add it to the same exception that is being used with valgrind

Signed-off-by: Carlo Marcelo Arenas Belón <redacted>
---
 t/test-lib.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Unlike the case for valgrind, where we actively do not want to set
these two environment variables [*1*], I am assuming that nedmalloc
simply ignores these two environment variables.  Is there a reason
why we want to special case nedmalloc like this patch does, but
leave these set for other implementations that do not pay attention
to them?  

Of course, I am also assuming that there are implementations of
malloc(3), which are not nedmalloc, that can be used to build and
test Git and that do not react to these two environment variables.
The patch would make sense if all the other implementations of
malloc(3) paid attention to MALLOC_{CHECK,PERTURB}_ and nedmalloc
were the only odd-man out, but I do not think that is the case.

[Footnote] 

*1* see the last two paragraphs of the log message of a731fa91 ("Add
MALLOC_CHECK_ and MALLOC_PERTURB_ libc env to the test suite for
detecting heap corruption", 2012-09-14)

quoted hunk
diff --git a/t/test-lib.sh b/t/test-lib.sh
index 44288cbb59..2ad9e6176c 100644
--- a/t/test-lib.sh
+++ b/t/test-lib.sh
@@ -143,7 +143,7 @@ fi
 # Add libc MALLOC and MALLOC_PERTURB test
 # only if we are not executing the test with valgrind
 if expr " $GIT_TEST_OPTS " : ".* --valgrind " >/dev/null ||
-   test -n "$TEST_NO_MALLOC_CHECK"
+   test -n "$TEST_NO_MALLOC_CHECK" || test -n "$USE_NED_ALLOCATOR"
 then
 	setup_malloc_check () {
 		: nothing
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help