- Dillo-dev - lists.mailman3.com

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

https related crash

by Johannes.Hofmann＠gmx.de

Hi, from time to time I get the crash below. Would a simple NULL check in a_Chain_check() be the correct fix? Cheers, Johannes nated with signal 11, Segmentation fault. #0 0x0805d359 in a_Chain_check (FuncStr=0x81135b3 "a_Capi_ccc", Op=2, Branch=1, Dir=2, Info=0x0) at chain.c:192 192 if (Info->Flags & (CCC_Ended + CCC_Aborted)) { (gdb) p Info $1 = (ChainLink *) 0x0 (gdb) bt #0 0x0805d359 in a_Chain_check (FuncStr=0x81135b3 "a_Capi_ccc", Op=2, Branch=1, Dir=2, Info=0x0) at chain.c:192 #1 0x08062f03 in a_Capi_ccc (Op=2, Branch=1, Dir=2, Info=0x0, Data1=0x28c83670, Data2=0x0) at capi.c:622 #2 0x0806372f in a_Capi_dpi_send_data (url=0x0, bw=0x2871da20, data=0x29563748 "<cmd='answer' to_cmd='dialog' msg='1' '>", data_sz=40, server=0x2940ee70 "proto.https", flags=<optimized out>) at capi.c:576 #3 0x0806381a in a_Capi_dpi_send_cmd (url=0x0, bw=0x2871da20, cmd=0x29563748 "<cmd='answer' to_cmd='dialog' msg='1' '>", server=0x2940ee70 "proto.https", flags=0) at capi.c:593 #4 0x08085bed in Dpiapi_dialog_answer_cb (answer=<optimized out>, bw=<optimized out>) at dpiapi.c:43 #5 a_Dpiapi_dialog (bw=0x2871da20, server=0x2940ee70 "proto.https", dpip_tag=0x286612d0 "<cmd='dialog' title='Dillo HTTPS: Missing issuer certificate!' msg='Unable to get local issuer certificate. The issuer certificate of an untrusted certificate cannot be found.' alt1='Continue' alt2='C"...) at dpiapi.c:74 #6 0x0805d21f in a_Chain_fcb (Op=684209776, Info=0x2878cec0, Data1=0x286612d0, Data2=0x29ecb9d0) at chain.c:114 #7 0x0808b533 in Dpi_parse_token (conn=<optimized out>) at dpi.c:228 #8 Dpi_process_dbuf (conn=<optimized out>, Data1=<optimized out>, Op=<optimized out>) at dpi.c:326 #9 a_Dpi_ccc (Op=2, Branch=2, Dir=1, Info=0x2878cec0, Data1=0x29e55470, Data2=0x0) at dpi.c:713 #10 0x0805d21f in a_Chain_fcb (Op=684209776, Info=0x28788fc0, Data1=0x29e55470, Data2=0x0) at chain.c:114 #11 0x0808ba6e in a_IO_ccc (Op=2, Branch=2, Dir=1, Info=0x28788fc0, Data1=0x292f6d40, Data2=0x0) at IO.c:425 #12 0x0808bc80 in IO_read (io=0x292f6d40) at IO.c:194 #13 0x0808bd88 in IO_callback (io=0x0) at IO.c:259 #14 0x0808be47 in IO_fd_read_cb (fd=6, data=0x6d7) at IO.c:280 #15 0x080e7d15 in fl_wait(double) () #16 0x080c36fc in Fl::wait(double) () #17 0x080c377b in Fl::run() () #18 0x0804e1f1 in main (argc=1, argv=0xbfbff8c4) at dillo.cc:502 (gdb)

Feb. 28, 2013

text-transform and hyphenation

by sgeerken＠dillo.org

On Sat, Feb 16, corvid wrote: > 'text-transform: capitalize' and hyphenation do not play well together > (example attached). The fragments are capitalized, and the word is > sometimes split as "Cap-" "Talization", losing the 'i'. Fixed. Sebastian

Feb. 16, 2013

text-transform and hyphenation

by corvid＠lavabit.com

'text-transform: capitalize' and hyphenation do not play well together (example attached). The fragments are capitalized, and the word is sometimes split as "Cap-" "Talization", losing the 'i'.

Feb. 16, 2013

form input button and margin

by sgeerken＠dillo.org

On Sun, Feb 02, corvid wrote: > I wrote: > > In the second one, the words and the button overlap. I'm not immediately > > sure what's special enough about that case to cause that. > > Here's some quick code made by looking at how the image code is working. > I would not be surprised if there are still complications lurking > somewhere in textblock, but it's at least something to play with... First of all, two remarks: (i) the border should also be added in getExtremesImpl; (ii) Embed::draw should either work similar to Image::draw (testing for intersection), or as currently, which is probably preferrably, since simpler, and drawing only some pixels more than needed. > Aside: It seems that, at least for firefox, the border is the edge of a > button or text entry widget or whatever, and padding goes inside that. > Of course we're not in any position to do that, but it's interesting. I do not have specs at hand, but I assume that this is correct. For this reason, I'd rather let the resource, not the embed widget, handle border and padding. For a first version, the standard borders (outset for button, inset for input fields, etc.) could be drawn; for an improvement, custom Fltk widgets would be neccessary. For reasons of uniformity, I'd let the resource als handle margin, which should then be implemented somewhere in flkt::ui. BTW: Johannes and I are planning to cleanup size handling; I'd like to get rid of size hints, and so also of the distinction "uses size hints". This will make it simpler to add <button>'s inline. Sebastian

Feb. 8, 2013

IO/proto.c

by p37sitdu＠lavabit.com

File is empty except for comment. Shouldn't it be removed?

Feb. 7, 2013

form input button and margin

by corvid＠lavabit.com

I wrote: > In the second one, the words and the button overlap. I'm not immediately > sure what's special enough about that case to cause that. Here's some quick code made by looking at how the image code is working. I would not be surprised if there are still complications lurking somewhere in textblock, but it's at least something to play with... Aside: It seems that, at least for firefox, the border is the edge of a button or text entry widget or whatever, and padding goes inside that. Of course we're not in any position to do that, but it's interesting.

Feb. 3, 2013