March 2013 - Dillo-dev - lists.mailman3.com

release preperations

by sgeerken＠dillo.org

Hi, I hope I am, as the developer of hyphenation, not biased on this topic. :-) > On Sat, Mar 02, 2013 at 07:57:26PM +0100, Sebastian Geerken wrote: > > > > As for the way to inform the user, see attached patch, which will let > > > > "make install" (and friends) print a message. Comments? > > > > > > Nice, but most users will get precompiled packages from their > > > distributions. > > > > > > > How is the other work on the release going on? What is still missing? > > > > > > I started a couple of times to add some lines to the man page that > > > explain how to get and install the pattern files, but never came up > > > with anything good. Should we copy the lines from your patch below? > > > > How many people compiling dillo from the tarball are actually reading > > README and the man page? > > At least the man page get's installed with most binary packages. > I tried to include some info about the pattern files and how > they can be downloaded - please give it a review. "man dillo" shows In order to use the hyphenation feature, pattern files from CTAN need to be installed to ${prefix}/lib/dillo/hyphenation/. This can be done with the script install-hyphenation. Call it with ISO-639-1 language codes as arguments, or without arguments to get more help. "${prefix}" should be replaced here. > Also we would need to install the install-hyphenation script, so > that it becomes part of binary packages. There would also some changes neccessary, since the script searches "Makefile". This is an aspect I have not yet thought about: whether the script is part of the installation process (as now) or the installation itself (as you suggest). > If we do that it would have to be renamed to > dillo-install-hyphenation or something like that. > Does this sound ok? I'd leave the question, how binary packages are handled, to the maintainers of the binary packages, since they know best how to handle this. (E. g., Debian provides packages texlive-lang-... for each language.) Perhaps package maintainers may comment on this. OTOH, if someone installs dillo from the tarball, and does not notice that she should download pattern files. Since hyphenation is inactive, when there is no pattern file installed, this feature will be invisible. That's why I thought to put it into the Makefile, so he sees the notice when he calls "make install". That's what I meant with my question. Sebastian

March 5, 2013

release preperations

by Johannes.Hofmann＠gmx.de

Hi, On Sat, Mar 02, 2013 at 07:57:26PM +0100, Sebastian Geerken wrote: > On Sat, Mar 02, Johannes Hofmann wrote: > > > I've pushed a new script, which uses the text versions (and adds the > > > licence information as comments). Should it download directly from > > > CTAN? From mirrors.dotsrc.org? Ask for a mirror? > > > > If mirrors.dotsrc.org works, that's probabely nicer - though I doubt > > we create any significant load. > > Mirrors.dotsrc.org works; I've changed the script. (Look at the > beginning.) > > > > As for the way to inform the user, see attached patch, which will let > > > "make install" (and friends) print a message. Comments? > > > > Nice, but most users will get precompiled packages from their > > distributions. > > > > > How is the other work on the release going on? What is still missing? > > > > I started a couple of times to add some lines to the man page that > > explain how to get and install the pattern files, but never came up > > with anything good. Should we copy the lines from your patch below? > > How many people compiling dillo from the tarball are actually reading > README and the man page? At least the man page get's installed with most binary packages. I tried to include some info about the pattern files and how they can be downloaded - please give it a review. Also we would need to install the install-hyphenation script, so that it becomes part of binary packages. If we do that it would have to be renamed to dillo-install-hyphenation or something like that. Does this sound ok? Cheers, Johannes PS: For me the documentation issue for hyphenation patterns is the only thing left before we could make a release candidate, or am I missing something?

March 5, 2013

https related crash

by jcid＠dillo.org

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.-

March 5, 2013

https related crash

by Johannes.Hofmann＠gmx.de

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

March 4, 2013

Profiling dillo

by p37sitdu＠lavabit.com

On Sat, Mar 02, 2013 at 11:36:16PM +0000, corvid wrote: > Sebastian wrote: > > (According to the man page, "gmon.out" is created in the current > > working directory, when exit() is called.) > > There's a chdir("/tmp") hiding in Paths::init(). > > Why don't remove it? >From ChangeLog: - Fixed Dillo and its dpis to work from "/tmp" (for easy device unmount). There was a thread discussing it: http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-December/009221.html http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-December/009223.html http://lists.auriga.wearlab.de/pipermail/dillo-dev/2011-December/009225.html Because of this, open dialog starts in /tmp, and it is a bug. Usually you start dillo from your home directory, why would you want to unmount it? And if you really plan to unmount it, you can change working directory yourself.

March 3, 2013

Profiling dillo

by corvid＠lavabit.com

Sebastian wrote: > (According to the man page, "gmon.out" is created in the current > working directory, when exit() is called.) There's a chdir("/tmp") hiding in Paths::init().

March 2, 2013

Profiling dillo

by sgeerken＠dillo.org

On Sat, Mar 02, Johannes Hofmann wrote: > On Fri, Mar 01, 2013 at 10:49:30AM +0100, Sebastian Geerken wrote: > > Has anyone tried to use gprof for profiling dillo? Using > > "--enable-gprof" for configure seems to work, and calling the tests > > (e. g.. test/dw-example) create the file "gmon.out", but not so > > does starting src/dillo. The option "-pg" is set everywhere, so I do > > not know what the difference is. > > > > Any ideas? > > Did you look in /tmp for the gmon.out file? Cores also end up there. Ah, yes, it is there. Thanks! (According to the man page, "gmon.out" is created in the current working directory, when exit() is called.) Sebastian

March 2, 2013

Profiling dillo

by Johannes.Hofmann＠gmx.de

Hi, On Fri, Mar 01, 2013 at 10:49:30AM +0100, Sebastian Geerken wrote: > Hi! > > Has anyone tried to use gprof for profiling dillo? Using > "--enable-gprof" for configure seems to work, and calling the tests > (e. g.. test/dw-example) create the file "gmon.out", but not so > does starting src/dillo. The option "-pg" is set everywhere, so I do > not know what the difference is. > > Any ideas? Did you look in /tmp for the gmon.out file? Cores also end up there. Johannes

March 2, 2013

https related crash

by jcid＠dillo.org

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.-

March 1, 2013

Profiling dillo

by sgeerken＠dillo.org

Hi! Has anyone tried to use gprof for profiling dillo? Using "--enable-gprof" for configure seems to work, and calling the tests (e. g.. test/dw-example) create the file "gmon.out", but not so does starting src/dillo. The option "-pg" is set everywhere, so I do not know what the difference is. Any ideas? Sebastian

March 1, 2013