Bug 217567
| Summary: | editors/emacs temacs dumps core during build on base/head amd64 using system compiler clang 4.0.1 | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Trond Endrestøl <Trond.Endrestol> | ||||||
| Component: | Individual Port(s) | Assignee: | Ashish SHUKLA <ashish> | ||||||
| Status: | Closed Overcome By Events | ||||||||
| Severity: | Affects Many People | CC: | jbeich, kib | ||||||
| Priority: | --- | Flags: | bugzilla:
maintainer-feedback?
(ashish) |
||||||
| Version: | Latest | ||||||||
| Hardware: | amd64 | ||||||||
| OS: | Any | ||||||||
| Attachments: |
|
||||||||
|
Description
Trond Endrestøl
2017-03-05 16:45:20 UTC
Can you still reproduce? head-* packages built fine on the cluster. http://www.ipv6proxy.net/go.php?u=http%3A%2F%2Fbeefy12.nyi.freebsd.org%2Fdata%2Flatest-per-pkg%2Femacs25%2F25.2_2%2C3%2Fhead-amd64-default.log&b=1 http://www.ipv6proxy.net/go.php?u=http%3A%2F%2Fbeefy11.nyi.freebsd.org%2Fdata%2Flatest-per-pkg%2Femacs25%2F25.2_2%2C3%2Fhead-i386-default.log&b=1 (In reply to Jan Beich from comment #1) It builds fine om base/head r320406. Most likely the attempts at fixing Stack Clash interfered for a while. Builds again on base/head r320406. A new but similar bug has appeared since last time.
Now I'm running base/head r321183 amd64, and temacs still fails to run during build of editors/emacs.
$ ./temacs
Fatal error 'Can't allocate initial thread' at line 337 in file /usr/src/lib/libthr/thread/thr_init.c (errno = 22)
Running ktrace reveals a failed call to break(2):
6355 temacs CALL break(0x409e11)
6355 temacs RET break -1 errno 22 Invalid argument
$ make showconfig
===> The following configuration options are available for emacs25-25.2_2,3:
ACL=off: Access control list support
CANNA=off: Japanese input support
DBUS=off: D-Bus IPC system support
GNUTLS=on: SSL/TLS support via GnuTLS
LLDB=on: LLDB support for GUD
MODULES=on: Dynamic modules support
SOURCES=on: Install sources
X11=on: X window system support
XML=on: XML format or parser support
====> Graphics support (each implies X11)
CAIRO=off: Experimental Cairo library support (implies PNG, XFT)
GIF=on: GIF image format support
JPEG=on: JPEG image format support
MAGICK=on: ImageMagick image processing support
PNG=on: PNG image format support
SVG=on: Scalable vector graphics support (implies PNG)
TIFF=on: TIFF image format support
XPM=on: XPM pixmap image format support
====> Other X11-dependent (each implies X11)
GCONF=off: GConf configuration backend support
GSETTINGS=off: GSettings API for application settings
M17N=on: M17N for text-shaping (implies XFT)
OTF=on: Opentype fonts (implies XFT)
SCROLLBARS=on: Toolkit scroll-bars
XFT=on: Xft font library support
XIM=on: X Input Method
XWIDGETS=off: GTK widgets in Emacs buffers (requires GTK3)
====> File notification support: you can only select none or one of them
GFILE=off: File notification via gfile
KQUEUE=on: File notification via kqueue
====> Sound (audio) support: you can only select none or one of them
ALSA=off: ALSA audio architecture support
OSS=off: Open Sound System support
====> X11 toolkit (each implies X11): you can only select none or one of them
GTK2=off: GTK+ 2 GUI toolkit support
GTK3=on: GTK+ 3 GUI toolkit support
MOTIF=off: Motif widget library support
XAW=off: Athena widgets
XAW3D=off: Athena3D widgets
===> Use 'make config' to modify these settings
Complete ktrace dump and typescript from running make build will be attached shortly.
Created attachment 184493 [details]
Complete ktrace dump from running temacs manually
Created attachment 184494 [details]
Complete
Comment on attachment 184494 [details]
Complete
Complete typescript from building editors/emacs. (Firefox is sure acting up, be it the backspace key or upwards on the scroll wheel on the mouse.)
EINVAL from break(2) means that the supplied address is less than the end of the data segment. Why break(2) was called there at all, and why it was passed the invalid argument, is the question to the application. Above were the facts that I see from the backtrace, now I am stating something that is a speculation, but based on my experience with debugging enough usermode code. I suspect that the program override the malloc(3) and the override is buggy, using obsolete break(2) interface. You need to check why it tries to do that and why it works otherwise. Why do you think that this is the same problem as the initial report, I do not know. (In reply to Konstantin Belousov from comment #8) The similar error messages led me astray. Anyway, editors/emacs and editors/emacs-nox11 works again at build and run time on base/head r321366 and ports/head r446377. Closed. |
