Pike-devel December 2012

pike-devel@lists.lysator.liu.se

3 participants
9 discussions

debugger opcodes
by Bill Welliver 30 Oct '14

30 Oct '14

I noticed that folks have been working on code generation lately. Any chance it might be possible to implement some of the opcodes required to get debugging working (step, break, etc)? I seem to recall someone indicating it wasn't a terribly difficult task, but it's clearly beyond my understanding of the problem. Bill

4 6

Sql.pgsql behavior in case of errors
by aldem-pikedev＠aldem.net 21 Aug '14

21 Aug '14

Hi everyone, I am preparing to use Postgres in my app, found the following strange behavior: ---snip--- Sql.Sql db = Sql.Sql("pgsql://localhost", "test", "test", "test"); db->query("BEGIN"); db->query("INSERT INTO requests (SvTrId) VALUES ('lala')"); catch(db->query("INSERT INTO requests (SvTrId) VALUES ('lala')")); db->query("INSERT INTO requests (SvTrId) VALUES ('juju')"); db->query("ROLLBACK"); ---snip--- Database is opened, row with 'lala' is inserted, second query (obviously) fails as SvTrId is the primary key, but row with 'juju' is inserted *and* appears in the database - i.e. the transaction is *committed*! Studying source code show, that once error is encounter, session is reset, i.e. running transaction is rolled back, everything is closed, temporary tables are deleted and session variables are reset as well, and after that, the next statement(s) will be committed immediately, as there is no active transaction. In my opinion, this is not exactly correct behavior when statement fails, as session still may be needed (especially temporary tables, open cursors etc). Im some cases, this is perfectly normal when SQL statement fails, especially in SERIALIZED transaction isolation level, and it would be more practical to leave the decision as of what to do to the application. Yes, this is true that it is impossible to continue executing statements in failed transaction, but temporary tables etc. may still be needed. Also, silent reconnection is not a good idea too - when there are many transactional statements are executed - it would be nice to have an option to modify this behavior. Best regards, /Alexander

5 6

Warnings for unused variables
by Jonas Walld�n ＠ Pike developers forum 31 Dec '12

31 Dec '12

When compiling Pike with clang I get massive amounts of warnings for unused variables. This detracts quite a bit from finding more serious errors and I've got a patch ready to remove them. However, before pushing that I'd like to hear if people have objections to changing declarations like in these examples: - static void exit_memory(struct object *o) + static void exit_memory(struct object *UNUSED(o)) and - static int got_udp_event (struct fd_callback_box *box, int event) + static int got_udp_event (struct fd_callback_box *box, int DEBUGUSED(event))

6 18

Filesystem.Tar()->get_dir() behaves oddly
by H. William Welliver III 20 Dec '12

20 Dec '12

I've been playing with a pluggable filesystem class that works with instances of Filesystem.Base, and have noticed that parts of Filesystem.Tar behave differently from Filesystem.System. In particular, paths expressed by Filesystem.Tar always seem to be full rather than the relative paths returned by Filesystem.System. That seems improper to me, especially if one expects Filesystem.Base()->get_dir() to behave in a manner similar to get_dir(). For example: > object f = Filesystem.Tar("foo.tar"); > f->get_dir(); (1) Result: ({ /* 2 elements */ "/Session.pmod", "/Tools.pmod" }) > f = f->cd("Session.pmod"); // note that the object is perfectly content to deal with relative paths > f->get_dir(); (3) Result: ({ /* 2 elements */ "/Session.pmod/RAMSessionStorage.pike", "/Session.pmod/SQLiteSessionStorage.pike", }) I've got a fix that seems to resolve this, but wanted to check that it's actually a bug (it seems as though it must be, as it would be hard for filesystem objects to be interchangeable if they behave differently in this way). Bill

1 0

Planned outage of pike.lysator.liu.se
by Henrik Grubbstr�m (Lysator) ＠ Pike (-) developers forum 18 Dec '12

18 Dec '12

It should be up now.

3 2

program names
by H. William Welliver III 11 Dec '12

11 Dec '12

I've got some code that generates pike program structs dynamically, and I was wondering if anyone could explain the mechanism that gives C-level programs their names when viewed from pike. For instance: object x = foo.MyProgram > object_program(x); (6) Result: dynamic_class.m:936 Where foo is a module which has `[] overloaded to generate c-api classes on demand. I assume that an entry would normally be created somewhere in the master, however because I'm bypassing all of that machinery, the necessary entries aren't being created. I assume that I should go back and create them, correct? If so, which tables really need to be populated? Bill

1 0

Assign where compare meant
by Peter Bortas ＠ Pike developers forum 03 Dec '12

03 Dec '12

Coverity picked up a few strange assignments in files/ that I commitet fixes for in 7.9. That they had been missed in the files module made me go hunting for more, and I found some in the signal handler. Does this make sense: diff --git a/src/signal_handler.c b/src/signal_handler.c index c886a86..b0a04bb 100644 --- a/src/signal_handler.c +++ b/src/signal_handler.c @@ -1076,7 +1076,7 @@ void forkd(int fd) do { j = close(i); } while ((j < 0) && (errno == EINTR)); - if ((j < 0) && (errno = EBADF)) num_fail++; + if ((j < 0) && (errno == EBADF)) num_fail++; } while (1) { @@ -1088,7 +1088,7 @@ void forkd(int fd) iov.iov_len = 1; do { i = recvmsg(fd, &msg, 0); - } while ((i < 0) && (errno = EINTR)); + } while ((i < 0) && (errno == EINTR)); if (!i) _exit(0); /* Connection closed, shutdown forkd. */ for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) { int ctrl_fd = NULL; @@ -1101,7 +1101,7 @@ void forkd(int fd) num_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int); do { i = fork(); - } while ((i < 0) && (errno = EINTR)); + } while ((i < 0) && (errno == EINTR)); if (i < 0) { /* Fork failure. */ } else if (i) { (Just in case those are there to acctually force the error to EINTR for some reason that I can't see.)

3 2

low_mega_apply switch
by Peter Bortas ＠ Pike developers forum 02 Dec '12

02 Dec '12

Is there a goot reason the default case for the big low_mega_apply switch is wedged in the middle and not at the end like normally done? If not I'm moving it.

1 0

Coverity Scan accounts
by Peter Bortas ＠ Pike developers forum 01 Dec '12

01 Dec '12

After dicussin it on the conferance I contacted Coverity and made a Pike project on their Open Source site. Those of you that want an account to check out and work on the defects it reports please reply to this with the user name and email address you want to use. If you don't want it in the list archives then send it to me directly at bortas(a)gmail.com.

1 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

Pike-devel December 2012