bjam' always 'bus error' on NetBSD/sparc (presumably, other strict-alignement 32 bits arch as well)

[rdolbeau]

Originally reported downstream at pkgsrc (http://gnats.netbsd.org/55269).

The observation:

Trying to compile the "boost" meta-package in pkgsrc on NetBSD 9.0/sparc, 'bjam' always crashes with a 'bus error'. The culprit is the function timestamp_init in tools/build/src/engine/timestamp.cpp, which does:

time->secs = secs;

where secs are time_t, currently 8 bytes. This is only legal on SPARC if time->secs is 8-bytes aligned, which it isn't, so the 8-bytes store causes a 'bus error'. The reason for the nonalignment is that it comes from the hash table implementation in hash.cpp, which does:

#define hash_item_data(item) ((HASHDATA *)((char *)item + sizeof(ITEM)))

ITEM is the size of a pointer, and item is properly aligned (at least 8 bytes), so this is never aligned properly with 32-bits pointers... This result in reliable crashes on 32-bits machine with strict alignment requirements (i.e., sparc).

my fix

The following patch fixes the problem for me, but might be a bit overkill (memory wastage) on 32-bits architecture that supports unaligned access. It should be a no-op on 64-bits architecture.

--- tools/build/src/engine/hash.cpp.orig        2019-12-10 01:20:17.000000000 +0100
+++ tools/build/src/engine/hash.cpp     2020-05-16 14:55:46.086210410 +0200
@@ -33,8 +33,8 @@
 typedef struct item ITEM;
 struct item
 {
-    ITEM * next;
-};
+    ITEM * next __attribute__ ((aligned (8)));
+} __attribute__ ((aligned (8)));

 #define MAX_LISTS 32

There is probably better solution than that, that would generally respect alinement's rules on all platforms.

[grafikrobot]

Thanks for reporting this.

[stale]

Thank you for your contributions. Main development of B2 has moved to https://github.com/bfgroup/b2
This issue has been automatically marked as "transition" to indicate the potential for needing transition to the new B2 development project.