June 2012 - Dillo-dev - lists.mailman3.com

signed integer overflow in cache.c

by jcid＠dillo.org

On Sat, Jun 16, 2012 at 01:31:01AM +0400, 123 wrote: > In function Cache_client_enqueue variable ClientKey is incremented and > checked for negative values after that (in case of overflow). > > Even though overflow is unlikely, signed integer overflow results in > wraparound on most systems and gcc don't seem to optimize away > overflow check even with -O3 (it only inlines Cache_client_enqueue in > my case) it is better to avoid overflow completely as there are other > compilers and compiler versions. +1. Agreed. I'd add a comment to remember this. e.g. /* integer wraparound/overflow is not defined by C/C++ standards */ -- Cheers Jorge.-

June 18, 2012

dillo.cc cleanup and bugfix

by jcid＠dillo.org

On Sat, Jun 16, 2012 at 02:49:47AM +0400, 123 wrote: > a_UIcmd_open_url and UIcmd_open_url_nbw are identical. Difference is > in argument name and comment. > > Why not remove UIcmd_open_url_nbw and replace it with > a_UIcmd_open_url? For simplicity. It is easier to differentiate the multiple situations: open in a new tab, in a new window, new bw (abstract). As a general rule we're not trying to optimize for size as the most important criteria these days; simplicity has served us better over the years. -- Cheers Jorge.-

June 18, 2012

dillo.cc cleanup and bugfix

by p37sitdu＠lavabit.com

a_UIcmd_open_url and UIcmd_open_url_nbw are identical. Difference is in argument name and comment. Why not remove UIcmd_open_url_nbw and replace it with a_UIcmd_open_url?

June 15, 2012

signed integer overflow in cache.c

by p37sitdu＠lavabit.com

And in case someone thinks it is non-issue: http://thiemonagel.de/2010/01/signed-integer-overflow/

June 15, 2012

signed integer overflow in cache.c

by p37sitdu＠lavabit.com

In function Cache_client_enqueue variable ClientKey is incremented and checked for negative values after that (in case of overflow). Even though overflow is unlikely, signed integer overflow results in wraparound on most systems and gcc don't seem to optimize away overflow check even with -O3 (it only inlines Cache_client_enqueue in my case) it is better to avoid overflow completely as there are other compilers and compiler versions.

June 15, 2012

dillo.cc cleanup and bugfix

by Johannes.Hofmann＠gmx.de

On Thu, Jun 14, 2012 at 11:37:46PM +0400, 123 wrote: > On Thu, Jun 14, 2012 at 07:00:17PM +0000, corvid wrote: > > Johannes wrote: > > > On Sun, Jun 10, 2012 at 10:39:21PM +0400, 123 wrote: > > > > aboutblank.diff removes unnecessary check. > > > > > Removed check for about:blank as start_page, it is not a special case. > > > > > > Committed. > > > > about:blank no longer gives focus to Location. > > Right, it is because a_UIcmd_open_url works differently for NULL and > about:blank. And it is commented in *identical* static function > UIcmd_open_url_nbw. > > I overlooked this difference. If you change it back, add comment > please. > done. Cheers, Johannes

June 15, 2012

Dillo working directory

by p37sitdu＠lavabit.com

And if we really want to change to unmounted directory, then why use "/tmp" instead of "/"? In this case all calls to fl_file_chooser should be changed to use old CWD.

June 15, 2012

Dillo working directory

by p37sitdu＠lavabit.com

On Sun, May 27, 2012 at 04:57:24PM +0400, 123 wrote: > In Paths::init Dillo changes working directory to /tmp. This way every > time I try to open some page stored on the hard disk (Ctrl-O) I have > to start in /tmp. What is the reason for changing WD to /tmp? BTW > oldWorkingDir is not used anywhere. oldWorkingDir is used in fact, it was a mistake. I have searched mailing list and found thread [1] [2] [3] discussing the same issue, but there is no rationale. I also found line in ChangeLog - Fixed Dillo and its dpis to work from "/tmp" (for easy device unmount). Most of the time Dillo is running from $HOME and unmounting it is not common. It may be that someone changes directory to, for example, /mnt, runs dillo there and exits shell (or changes its WD), but it is not common scenario too. The problem with changing directory to /tmp is that open dialog starts in /tmp, it is a bug. As a workaround I am typing "file:" in location, it uses oldWorkingDir then. I have attached patch, it fixes the problem. It is still not finished without removing obsolete oldWorkingDir and using real current directory. Are there any other hidden problems? If not, I will make finished patch. [1] http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-December/009221.html [2] http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-December/009223.html [3] http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-December/009225.html

June 14, 2012

dillo.cc cleanup and bugfix

by p37sitdu＠lavabit.com

On Thu, Jun 14, 2012 at 07:00:17PM +0000, corvid wrote: > Johannes wrote: > > On Sun, Jun 10, 2012 at 10:39:21PM +0400, 123 wrote: > > > aboutblank.diff removes unnecessary check. > > > > Removed check for about:blank as start_page, it is not a special case. > > > > Committed. > > about:blank no longer gives focus to Location. Right, it is because a_UIcmd_open_url works differently for NULL and about:blank. And it is commented in *identical* static function UIcmd_open_url_nbw. I overlooked this difference. If you change it back, add comment please.

June 14, 2012

dillo.cc cleanup and bugfix

by Johannes.Hofmann＠gmx.de

On Thu, Jun 14, 2012 at 07:00:17PM +0000, corvid wrote: > Johannes wrote: > > On Sun, Jun 10, 2012 at 10:39:21PM +0400, 123 wrote: > > > aboutblank.diff removes unnecessary check. > > > > Removed check for about:blank as start_page, it is not a special case. > > > > Committed. > > about:blank no longer gives focus to Location. > ah, the details... I reverted it.

June 14, 2012