Hi, from time to time I get the crash below. Would a simple NULL check in a_Chain_check() be the correct fix? Cheers, Johannes nated with signal 11, Segmentation fault. #0 0x0805d359 in a_Chain_check (FuncStr=0x81135b3 "a_Capi_ccc", Op=2, Branch=1, Dir=2, Info=0x0) at chain.c:192 192 if (Info->Flags & (CCC_Ended + CCC_Aborted)) { (gdb) p Info $1 = (ChainLink *) 0x0 (gdb) bt #0 0x0805d359 in a_Chain_check (FuncStr=0x81135b3 "a_Capi_ccc", Op=2, Branch=1, Dir=2, Info=0x0) at chain.c:192 #1 0x08062f03 in a_Capi_ccc (Op=2, Branch=1, Dir=2, Info=0x0, Data1=0x28c83670, Data2=0x0) at capi.c:622 #2 0x0806372f in a_Capi_dpi_send_data (url=0x0, bw=0x2871da20, data=0x29563748 "<cmd='answer' to_cmd='dialog' msg='1' '>", data_sz=40, server=0x2940ee70 "proto.https", flags=<optimized out>) at capi.c:576 #3 0x0806381a in a_Capi_dpi_send_cmd (url=0x0, bw=0x2871da20, cmd=0x29563748 "<cmd='answer' to_cmd='dialog' msg='1' '>", server=0x2940ee70 "proto.https", flags=0) at capi.c:593 #4 0x08085bed in Dpiapi_dialog_answer_cb (answer=<optimized out>, bw=<optimized out>) at dpiapi.c:43 #5 a_Dpiapi_dialog (bw=0x2871da20, server=0x2940ee70 "proto.https", dpip_tag=0x286612d0 "<cmd='dialog' title='Dillo HTTPS: Missing issuer certificate!' msg='Unable to get local issuer certificate. The issuer certificate of an untrusted certificate cannot be found.' alt1='Continue' alt2='C"...) at dpiapi.c:74 #6 0x0805d21f in a_Chain_fcb (Op=684209776, Info=0x2878cec0, Data1=0x286612d0, Data2=0x29ecb9d0) at chain.c:114 #7 0x0808b533 in Dpi_parse_token (conn=<optimized out>) at dpi.c:228 #8 Dpi_process_dbuf (conn=<optimized out>, Data1=<optimized out>, Op=<optimized out>) at dpi.c:326 #9 a_Dpi_ccc (Op=2, Branch=2, Dir=1, Info=0x2878cec0, Data1=0x29e55470, Data2=0x0) at dpi.c:713 #10 0x0805d21f in a_Chain_fcb (Op=684209776, Info=0x28788fc0, Data1=0x29e55470, Data2=0x0) at chain.c:114 #11 0x0808ba6e in a_IO_ccc (Op=2, Branch=2, Dir=1, Info=0x28788fc0, Data1=0x292f6d40, Data2=0x0) at IO.c:425 #12 0x0808bc80 in IO_read (io=0x292f6d40) at IO.c:194 #13 0x0808bd88 in IO_callback (io=0x0) at IO.c:259 #14 0x0808be47 in IO_fd_read_cb (fd=6, data=0x6d7) at IO.c:280 #15 0x080e7d15 in fl_wait(double) () #16 0x080c36fc in Fl::wait(double) () #17 0x080c377b in Fl::run() () #18 0x0804e1f1 in main (argc=1, argv=0xbfbff8c4) at dillo.cc:502 (gdb)
Hi Johannes, On Thu, Feb 28, 2013 at 10:44:15PM +0100, Johannes Hofmann wrote:
Hi,
from time to time I get the crash below. Would a simple NULL check in a_Chain_check() be the correct fix?
I gave it a first review, and it looks like the connection gets closed before the dialog sends an answer. So, at resume time, the sending branch is gone. It'd be great to have a test case. Anyway, in the interim, please try the following patch: diff -r 8c8975054b06 src/capi.c --- a/src/capi.c Wed Jan 30 10:04:00 2013 +0100 +++ b/src/capi.c Fri Mar 01 13:32:38 2013 -0300 @@ -571,10 +571,13 @@ int a_Capi_dpi_send_data(const DilloUrl /* Re-use an open connection */ conn = Capi_conn_find(server); if (conn) { - /* found */ - dbuf = a_Chain_dbuf_new(data, data_sz, 0); - a_Capi_ccc(OpSend, 1, BCK, conn->InfoSend, dbuf, NULL); - dFree(dbuf); + if (conn->InfoSend) { + /* found & operative*/ + dbuf = a_Chain_dbuf_new(data, data_sz, 0); + a_Capi_ccc(OpSend, 1, BCK, conn->InfoSend, dbuf, NULL); + dFree(dbuf); + } else + MSG(" ERROR: [a_Capi_dpi_send_data] Connection not operative\n"); } else { MSG(" ERROR: [a_Capi_dpi_send_data] No open connection found\n"); } diff -r 8c8975054b06 src/chain.c --- a/src/chain.c Wed Jan 30 10:04:00 2013 +0100 +++ b/src/chain.c Fri Mar 01 13:32:38 2013 -0300 @@ -189,7 +189,10 @@ int a_Chain_check(char *FuncStr, int Op, /* Show status information */ Chain_debug_msg(FuncStr, Op, Branch, Dir, Info); - if (Info->Flags & (CCC_Ended + CCC_Aborted)) { + if (!Info) { + MSG_WARN("CCC: call on a NULL node.\n" + "Caught as last resort; Most probably a BUG.\n"); + } else if (Info->Flags & (CCC_Ended + CCC_Aborted)) { /* CCC is not operative */ MSG_WARN("CCC: call on already finished chain. Flags=%s%s\n", Info->Flags & CCC_Ended ? "CCC_Ended " : "", -- Cheers Jorge.-
On Fri, Mar 01, 2013 at 01:37:13PM -0300, Jorge Arellano Cid wrote:
Hi Johannes,
On Thu, Feb 28, 2013 at 10:44:15PM +0100, Johannes Hofmann wrote:
Hi,
from time to time I get the crash below. Would a simple NULL check in a_Chain_check() be the correct fix?
I gave it a first review, and it looks like the connection gets closed before the dialog sends an answer. So, at resume time, the sending branch is gone. It'd be great to have a test case.
Sorry, I can't reproduce it atm. I keep trying.
Anyway, in the interim, please try the following patch:
Will do, once I can reproduce the crash. Thanks, Johannes
On Mon, Mar 04, 2013 at 10:07:21PM +0100, Johannes Hofmann wrote:
On Fri, Mar 01, 2013 at 01:37:13PM -0300, Jorge Arellano Cid wrote:
Hi Johannes,
On Thu, Feb 28, 2013 at 10:44:15PM +0100, Johannes Hofmann wrote:
Hi,
from time to time I get the crash below. Would a simple NULL check in a_Chain_check() be the correct fix?
I gave it a first review, and it looks like the connection gets closed before the dialog sends an answer. So, at resume time, the sending branch is gone. It'd be great to have a test case.
Sorry, I can't reproduce it atm. I keep trying.
Anyway, in the interim, please try the following patch:
Will do, once I can reproduce the crash.
AFAIR there're a few obscure corner cases not explicitly handled by the CCC (ATM not worth the coding effort), and those are silently caught by the CCC's check code. From a distance, and with the proposed patch, it doesn't look like a show stopper at all. -- Cheers Jorge.-
participants (2)
-
jcid＠dillo.org -
Johannes.Hofmann＠gmx.de