Hinn 9. ágúst 2009 lét Aurelien Jarno þetta frá sér fara: > On Sun, Aug 09, 2009 at 05:22:27PM +0000, Bjarni Ingi Gislason wrote: > > Some more information about dump in ftp. > > > > 1) No dump if text is keyed in, instead of using the macro "$safn ...". > > (only tested once) > > Is the problem always reproducible? >
No, but I found a solution, see next item. ... > > dir std* "| more" > > 200 PORT command successful. > > 150 Opening ASCII mode data connection for /bin/ls. > > *** glibc detected *** ftp: corrupted double-linked list: 0x09ceb050 *** > > This sounds like a bug in the ftp package. > Thanks for the tip. This bug has been reported for the ftp package. Bug #508378 explains the situation. One then only needs to add enough spaces to the definition of the macro, in my case 2 spaces after the first word, to get it working. I think I should report the segmentation fault from "ace-of-penguins" to that package. There are still issues with "malloc.c" on my mind. 1) The libc6-dbg package should not be compiled with the -O option (see "<value optimized out>" in a backtrace). Use that library only for the program to be tested (e.g. env LD_LIBRARY_PATH=/usr/lib/debug <program(s) ...>. or add the name after ".../debug" if the variable exists. Put that in a file and source the file. Repeating is thus simple. Such information could be in a README file (now absent) with some references about debugging. 2) The segmentation fault is in the "unlink(nextchunk, bck, fwd)" function, but it is a "#define". Therefore no backtrace??? Maybe a compile dependent version of "unlink" should be used if debugging information is created (option -g). 3) Even an "malloc(512)" can cause a segmentation fault (example with "mtrace()", line 314: malloc (TRACE_BUFFER_SIZE) ). 4) Errors (warnings) could be better phrased for the novices Suggestion: Instead of *** glibc detected *** ftp: corrupted double-linked list: 0x09ceb050 *** use e.g. *** glibc detected invalid data from ftp: ... Generally state with a sentence (if known): 1) Who is reporting 2) What is done 3) Why is it done (cause of problem) 4) Who is the culprit -- Bjarni I. Gislason -- To UNSUBSCRIBE, email to debian-glibc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org