ROE test fails on Debian unstable/amd64
By Thomas Girard - Posted on March 17th, 2008
Tagged:
| Project: | GNU Smalltalk |
| Component: | Testsuite |
| Category: | bug |
| Priority: | normal |
| Assigned: | Unassigned |
| Status: | fixed |
Description
Hello,
test 117 of GNU Smalltalk 3.0a fails on Debian amd64 sid box. (It also fails on i386.)
Please find attached the testsuite.log associated with this test. I seem to recall there was an API change in SQLite3 (we have 3.5.6), it might be related.
I'll have a look at this if you don't.
Updates
#1 submitted by Paolo Bonzini on Tue, 03/18/2008 - 09:21
Interesting, I see it too with 3.5.6 (not with 3.4.2). This patch fixes it, I think. It's a double free or something like that:
diff --git a/packages/dbd-sqlite/sqlite3.c b/packages/dbd-sqlite/sqlite3.c index fda8390..55d34ba 100644 --- a/packages/dbd-sqlite/sqlite3.c +++ b/packages/dbd-sqlite/sqlite3.c @@ -217,8 +217,11 @@ gst_sqlite3_finalize (OOP self) SQLite3StmtHandle h; h = (SQLite3StmtHandle) OOP_TO_OBJ (self); - stmt = (sqlite3_stmt *) vmProxy->OOPToCObject (h->stmt); + if (h->stmt == vmProxy->nilOOP) + return 0; + stmt = (sqlite3_stmt *) vmProxy->OOPToCObject (h->stmt); + h->stmt = vmProxy->nilOOP; return sqlite3_finalize (stmt); }
#2 submitted by Thomas Girard on Tue, 03/18/2008 - 10:07
| Status: | active | » committed |
Hello, your patch fixes the failure here too :-)
Assuming `fixed' status is for bugs fixes that are available in a release, I'm marking this bug `committed'.
#3 submitted by Paolo Bonzini on Tue, 03/18/2008 - 10:19
| Status: | committed | » fixed |
Wrong assumption, committed is for fixes that are in someone's git repository (so that I can just cherrypick them). :-)