Pike-devel September 2014

pike-devel@lists.lysator.liu.se

9 participants
28 discussions

testsuite
by Martin Nilsson (Opera Mini - AFK!) ＠ Pike (-) developers forum 17 Sep '14

17 Sep '14

Recent changes to how the testsuite is run makes error messages from expected failures to be printed out. That is not helpful. Also, we've gone from 0 to 42 failing tests.

6 23

testsuite
by Per Hedbor () ＠ Pike (-) developers forum 17 Sep '14

17 Sep '14

Sorry, I had an error in my "it will now return" mapping. Actual example: > p->finish("<t a='b''c'\"d\"e>")->read(); ([ /* 3 elements */ "a": "b", "c": "c", "d": "d" "e": "e" ])

4 7

lfun `+= and bogus conditions, performance loss, semantic errors
by Stephen R. van den Berg 17 Sep '14

17 Sep '14

Object `+= lfuns are not called when they should have been. It appears that we don't call += lfuns directly, instead we expand a += b+c into a = a+b+c. Well, that is all fine and dandy, but, the f_add function then tries to fold this back into a += by checking if the number of refs on the object == 1, and then calling `+= instead. This check does not give any meaningful results given normal objects and += usage: - For one, the object might have an extra instance on the stack, in which case the minimum number of references becomes 2 already. - And for another, the object might have multiple instances which all refer to the same object; which *implies* that updating one of those using += should modify all of them. The following sample code: class refmeharder { int val; void create(int i) { werror("Set %d\n",val = i); }; void getval(string s) { werror("%s%d\n",s,val); }; void incval() { werror("Inc %d\n",++val); }; refmeharder `+=(int i) { werror("+= called %d\n", val+=i); return this; }; refmeharder `+(int i) { werror("+ called %d\n", val+i); return refmeharder(val+i); }; }; int main(int argc, array(string) argv) { refmeharder a=refmeharder(2); refmeharder b=refmeharder(3); refmeharder c=a,d=a,e=a,f=a,g=a,h=a; a->incval(); b->incval(); c->incval(); a+=5; b+=11; c+=23; a->getval("a"); b->getval("b"); c->getval("c"); return 0; } Together with the following debug-patch: diff --git a/src/operators.c b/src/operators.c index 45de79f..18fbe4a 100644 --- a/src/operators.c +++ b/src/operators.c @@ -1505,6 +1505,8 @@ PMOD_EXPORT void f_add(INT32 args) /* The first argument is an object. */ o = sp[-args].u.object; p = o->prog->inherits[SUBTYPEOF(sp[-args])].prog; +if((i = FIND_LFUN(p, LFUN_ADD_EQ)) != -1) +fprintf(stderr, "Object %p type %p refs %d\n", o, p, o->refs); if(o->refs==1 && (i = FIND_LFUN(p, LFUN_ADD_EQ)) != -1) { Results in an output of: Set 2000 Set 3000 Inc 2001 Inc 3001 Inc 2002 Object 0x200e750 type 0x2055450 refs 8 + called 2007 Set 2007 Object 0x200e7b0 type 0x2055450 refs 2 + called 3012 Set 3012 Object 0x200e750 type 0x2055450 refs 7 + called 2025 Set 2025 a2007 b3012 c2025 But should have resulted in an output of: Set 2000 Set 3000 Inc 2001 Inc 3001 Inc 2002 Object 0x200e750 type 0x2055450 refs 8 += called 2007 Object 0x200e7b0 type 0x2055450 refs 2 += called 3012 Object 0x200e750 type 0x2055450 refs 7 += called 2040 a2040 b3012 c2040 I'll be adding something to this effect to the testsuites. The immediate questions though here are: - The refs==1 check is way too limiting, the question is, is there a meaningful alternative? (Probably not). - The only way to solve this seems to be by supporting calling `+= lfuns directly. But, if we do, do we need to somehow suppress the autoconversion from a+=b+c to a=a+b+c ? Can it be suppressed for objects only? Where in the code is this conversion being done? -- Stephen.

10 51

IOBuffer usability
by Martin Nilsson (Opera Mini - AFK!) ＠ Pike (-) developers forum 17 Sep '14

17 Sep '14

I changed part of the ASN1 decoder to use IOBuffer. Some comments - Stdio.IOBuffer is harder to type than ADT.struct. Can we rename it to Stdio.Buffer? There is already IO in Stdio. - get_ is shorter than read_. I'm on the fence on if this is a good change though. - read_buffer dosn't return Stdio.IOBuffer. An expression of type { IOBuffer = object } cannot be assigned to a variable of type { IOBuffer = object(implements _static_modules._Stdio()->IOBuffer) }. - It would be great if we could type cast() properly, so we can tell that (string)iobuffer is actually string(8bit).

2 3

Linux 3.13
by Martin Nilsson (Opera Mini - AFK!) ＠ Pike (-) developers forum 17 Sep '14

17 Sep '14

I've upgraded to Linux 3.13.0-34 and now the async_tls_close_test doesn't work anymore when called with 0 1 as parameters (blocking without clean close). Any ideas what to look for?

4 13

Refcounting gone wild?
by Stephen R. van den Berg 17 Sep '14

17 Sep '14

Is this to be expected? Why does it run up? > object c=Stdio.File(); > _refs(c); Result: 2 > c=c; Result: Stdio.File(0, 0, 777 /* fd=-1 */) > _refs(c); Result: 3 > c=c; Result: Stdio.File(0, 0, 777 /* fd=-1 */) > _refs(c); Result: 4 > -- Stephen.

5 4

IOBuffer vs. Buffer
by Stephen R. van den Berg 17 Sep '14

17 Sep '14

The actual reason I beefed up Buffer a bit lately is *because* I need to do some protocol decoding of a byte stream. Now I see IOBuffer arrive. In order to avoid code bloat, wouldn't it be a better idea to integrate the functionality of IOBuffer into Buffer and just keep one? Or is the performance difference so strikingly great that this sounds like a bad idea? -- Stephen.

5 16

configure tests
by Martin Nilsson (Opera Mini - AFK!) ＠ Pike (-) developers forum 17 Sep '14

17 Sep '14

Would it be reasonable to stop testing for functions defined in C89?

2 4

Does this fix a real bug?
by Stephen R. van den Berg 17 Sep '14

17 Sep '14

While "browsing" some code (what else to do with your spare time ;-), I find (not committed yet): commit e8716b0a9313e5c32b4dbf1d9c6246cad5a6b3b2 Author: Stephen R. van den Berg <srb(a)cuci.nl> Date: Mon Sep 1 11:31:36 2014 +0200 Store new malloced value instead. diff --git a/src/stralloc.c b/src/stralloc.c index 7cd709c..6ca0a02 100644 --- a/src/stralloc.c +++ b/src/stralloc.c @@ -3221,10 +3221,8 @@ PMOD_EXPORT void free_string_builder(struct string_builder *s) PMOD_EXPORT struct pike_string *finish_string_builder(struct string_builder *s) { ptrdiff_t len = s->s->len; - if (len != s->malloced) { - s->s->len = s->malloced; - s->s = realloc_unlinked_string(s->s, len); - } + if (len != s->malloced) + s->s = realloc_unlinked_string(s->s, s->malloced = len); else /* Ensure NUL-termination */ low_set_index(s->s,s->s->len,0); Am I just reading this wrong, or was the len/malloced assignment the wrong way around? And if yes, does this have any noticeable impact on the rest of the system? -- Stephen.

2 1

`<= vs `==
by Marcus Comstedt (ACROSS) (Hail Ilpalazzo!) ＠ Pike (-) developers forum 17 Sep '14

17 Sep '14

Either that, or add ``<, but then you haven't actually won anything.

1 0

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 September 2014