Segmentation fault (core dumped) due to missing write permissions

[Hermann Kleier]

I am using “kicad 0.201602050916+6538~42~ubuntu15.10.1”. I record my *.kicad_mod files with cvs. That makes my footprint read-only unless I explicitly invoke “cvs edit”. Every time when I try to overwrite my footprint-files with the footprint editor without making the file writeable, kicad aborts with the given message. I am aware that this is a plain user error. An error message telling me that the write permission is missing would be helpful. I reckon that this problem is straightforward. Would it be helpful to send the core file for more complex errors?

[Nick Østergaard (nickoe)]

A backtrace from the coredump will be enough I think.

I can't reproduce the segfault though. What I did was chmod -w on a kicad_pcb file and opened it with pcbnew. Can you state the exact steps to reproduce, and maybe without cvs?

[Nick Østergaard (nickoe)]

... and of course modify the file and try to save it. (I did it on both the kicad_pcb and pro file)

[Chris Pavlina (pavlina-chris)]

Yup, I get "You do not have write permissions to save file </home/cmp/noname.kicad_pcb>." here. Can you give more exact steps to reproduce the problem? Could be something subtle.

[Chris Pavlina (pavlina-chris)]

To clarify, it's obvious that I cannot read, but when I try on a footprint file instead...

"IO_ERROR: user does not have write permission to delete file '/home/cmp/git/efm8ub2/pcblib/conn-cui.pretty/CONN-PJ-037A.kicad_mod'
from kicad_plugin.cpp : FootprintSave() : line 1853"

No segfault. This is a debug build, I don't have a release build handy - maybe it behaves differently there?

[pipe (pipatron)]

When I do this in Debian Testing I get a reasonably nice error requester saying:

IO_ERROR: user does not have write permission to delete file '/home/pipe/abc.pretty/abc.kicad_mod'
from kicad_plugin.cpp : FootprintSave() : line 1853

or if I write-protect abc.pretty:

IO_ERROR: Library '/home/pipe/abc.pretty' is read only
from kicad_plugin.cpp : FootprintSave() : line 1834

Clicking "Ok" or whatever, takes me back to the footprint editor without any issues. Maybe because I use a debug build?

Application: pcbnew
Version: (2016-02-02 BZR 6535, Git 0ef1eee)-product debug build
wxWidgets: Version 3.0.2 (debug,wchar_t,compiler with C++ ABI 1009,GCC 5.3.1,wx containers,compatible with 2.8)
Platform: Linux 4.3.0-1-amd64 x86_64, 64 bit, Little endian, wxGTK
Boost version: 1.58.0
         USE_WX_GRAPHICS_CONTEXT=OFF
         USE_WX_OVERLAY=OFF
         KICAD_SCRIPTING=ON
         KICAD_SCRIPTING_MODULES=ON
         KICAD_SCRIPTING_WXPYTHON=ON
         USE_FP_LIB_TABLE=HARD_CODED_ON
         BUILD_GITHUB_PLUGIN=ON

[Hermann Kleier (hermann-kleier)]

Sorry for my unclear report. The procedure to observe the problem is:

 - Have the kicad_mod file r/o.
 - Start the footprint editor from the kicad main menu and edit a footprint.
 - Use the “Export footprint” button and navigate to the r/o file.
 - Click and bomb.

One of my tests (today) opened a window asking me whether I wanted to send a report. I agreed. So may be you get some more info from that.

Attached you find the stack of the core.

[Maciej Suminski (orsonmmz)]

I am sorry, I must have marked revision 6543 as the bug fix by mistake. If there were no other changes to fix it, then the bug must be still present.

[Hermann Kleier (hermann-kleier)]

Hi Maciej,

thank you for making it clear. I’ll wait for the next revision.

Hermann

On Tue, 9 Feb 2016, Maciej Sumiński wrote:

> I am sorry, I must have marked revision 6543 as the bug fix by mistake.
> If there were no other changes to fix it, then the bug must be still
> present.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1542856
>
> Title:
> Segmentation fault (core dumped) due to missing write permissions
>
> Status in KiCad:
> Fix Committed
>
> Bug description:
> I am using “kicad 0.201602050916+6538~42~ubuntu15.10.1”. I
> record my *.kicad_mod files with cvs. That makes my footprint read-
> only unless I explicitly invoke “cvs edit”. Every time when I try to
> overwrite my footprint-files with the footprint editor without making
> the file writeable, kicad aborts with the given message. I am aware
> that this is a plain user error. An error message telling me that the
> write permission is missing would be helpful. I reckon that this
> problem is straightforward. Would it be helpful to send the core file
> for more complex errors?
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/kicad/+bug/1542856/+subscriptions
>

[Hermann Kleier (hermann-kleier)]

I’ve upgraded to revision 6545. The problem is still there (see attachment).

Procedure used:

 - Start kicad from the command line.
 - From there: start the footprint editor using a read-only footprint
 - Edit the footprint
 - Click the “Export Footprint” button
 - Select the footprint-file (read-only!)
 - Click “Save”
 - The warning “A file named "CUS-12B.kicad_mod" already exists. Do you want to replace it? …” is displayed.
 - Click “Replace”
 - There is no error message then. The program abort immediately.
 - The console reads: “Segmentation fault (core dumped)”.

[jean-pierre charras (jp-charras)]

I tried to do that, and I have no crash, just the message about not writable file.

[Hermann Kleier (hermann-kleier)]

Meanwhile I am using build 6552. I have tried to isolate the problem as far as possible. I made up a fresh directory, placed an empty footprint file in it, made it read-only, and started kicad. All the steps are documented in the protocol which you can find in the attached file. All the steps within kicad are documented in the screenshots p01.png … p12.png which are recorded in chronological order. The “Replace”-confirmation was the last thing I could see from kicad.

I recorded the command line steps from an emacs-shell. Actually, I had observed the problem initially when I started kicad from the bare command line. To verify that it is not an emacs-artifact I performed the same steps from the command line —under X11, of course—. → same result.

To verify that the problem is not related to my current project file (th433.pro) I started by creating a new project and continued with the described steps. → same result.

To make sure that the problem does not depend on some strange setup on my workstation or even a hardware bug I installed kicad on a computer that had never seen kicad before and tried the very same procedure. Just like my workstation it was Ubuntu 15.10. → Same results.

I attach another file (bug_virgin) from that computer. There was the problem that I never had done any project on this device and therefore the footprint paths had not been set up before. I ignored the warning (see /home/h/Pictures/t01.png) and proceeded as explained before. Though the message “Segmentation fault (core dumped)” was written there was no core because cores are suppressed on Ubuntu by default (and I did not change that on that machine).

The warning showed me that all that could depend on the footprint paths in OTHER projects. There I had replaced the ”/usr/share/…” path by project-local paths because I wanted to keep track of them (cvs).

How can I provide more information?

[Hermann Kleier (hermann-kleier)]

The missing attachments…

[Hermann Kleier (hermann-kleier)]

Sorry, but I fail to confirm that the problem is gone. Meanwhile I have upgraded to kicad 0.201602211417+6576~42~ubuntu15.10.1.

Today I tried three cases: The isolated case with nothing more than a write-protected file and a real-world case. I got three different results.

 1 The first case happens most likely when exporting footprints to w/o-files: A null pointer is passed to __GI__IO_fputs. (see kicad_core)

 2 The case showing the “error report window” is rare. It happened from the real-world project when trying to overwrite a read-only file. A core file was written too. (see kicad_core.png)

 3 One time an “error window” opened telling me that the footprint file is write-protected. This is the good case. Yes, it happened once!

 4 Finally I observed a strange case. I copied a set of 25 footprint-files to an empty directory and put an empty w/o-file in it (named dummy.kicad_mod). I configured the new directory as a library directory (the only one) and tried to overwrite the empty w/o-file. I got a parse-error. Repeating the attempt to store gave me another (different!) parse error. (see parse_error.png and parse_error_2.png). This could be repeated again and again. But when I tried to export my dummy pattern to the w/o-file, I got the core.

[jean-pierre charras (jp-charras)]

If the crash happens only for the "export footprint" command, not the "save footprint", it is fixed in rev 6578.
It is Linux specific.

I did not seen any crash with the "save footprint" command.

[Hermann Kleier (hermann-kleier)]

Okay, that sounds good. I’ll wait for rev 6578 and rerun the tests.

Hermann

[Nick Østergaard (nickoe)]

The commit made for this fix was cherry picked to 4.0.2.

[Hermann Kleier (hermann-kleier)]

Fix confirmed in “kicad 0.201603061448+6610~43~ubuntu15.10.1”. Thank you.