bugmake - Bugs: bug #20033, parallel (-j2) make with $(eval)...

 
 

bug #20033: parallel (-j2) make with $(eval) construct segfaults

Submitted by:  None
Submitted on:  Fri 01 Jun 2007 10:48:41 AM UTC  
 
Severity: 3 - NormalItem Group: Bug
Status: FixedPrivacy: Public
Assigned to: Paul D. Smith <psmith>Open/Closed: Closed
Component Version: 3.81Operating System: POSIX-Based
Fixed Release: 3.82Triage Status: None

Add a New Comment(Rich Markup)
   

You are not logged in

Please log in, so followups can be emailed to you.

 

Thu 14 Jun 2007 12:41:25 PM UTC, comment #5:

Thanks, that works for me and mine.

Cheers,
Rasmus

Rasmus Andersen <rasmus>
Thu 14 Jun 2007 02:27:57 AM UTC, comment #4:

OK, I found it. I'm still not 100% sure why it makes a difference between parallel and non-parallel jobs but I think I understand it. The attached patch definitely seems to fix the problem.

(file #13043)

Paul D. Smith <psmith>
Project AdministratorIn charge of this item.
Thu 14 Jun 2007 01:29:56 AM UTC, comment #3:

I can reproduce this as well and the fix seems simple enough. The thing that's concerning me is WHY does it only happen with parallel makes? There must be something non-obvious happening here because it doesn't seem to me like parallelism should change this behavior in the least.

I need to look at that before I'll feel comfortable just fixing the problem.

Paul D. Smith <psmith>
Project AdministratorIn charge of this item.
Fri 01 Jun 2007 08:43:52 PM UTC, comment #2:

(I can reproduce a crash with that example.)

Martin Dorey <mdorey>
Fri 01 Jun 2007 07:52:33 PM UTC, comment #1:

Hi, its me again. Find below a trivial makefile to recreate this. run 'make setup && make -j2' to get a segfault.

Cheers,
Rasmus

A := $(wildcard *.a)

.PHONY: all
all: $(A:.a=.b)

.PHONY: clean
clean:
rm -f $(A)

.PHONY: setup
setup:
@touch 1.a 2.a 3.a 4.a 5.a 6.a

define d1
@echo lala $(1)
@sleep 1
endef

define d2
@echo tyty $(1)
@sleep 1
endef

%.b : %.a
$(eval CHECKSUM := $(word 1,$(shell cat $^ | sha1sum))) $(if $(wildcard $(CACHEDIR)/$(CHECKSUM)),\
$(call d1,$(CHECKSUM)),\
$(call d2,$(CHECKSUM)))

Anonymous
Fri 01 Jun 2007 10:48:41 AM UTC, original submission:

Hello,

At work I have a (proprietary, sorry) makefile where I use an $(eval) construct. It goes like

foo: bar
$(eval CHECKSUM := $(word 1,$(shell cat $^ | sha1sum))) $(if $(wildcard $(CACHEDIR)/$(CHECKSUM)), $(call do-copy,$(CHECKSUM)), $(call do-dct4-selection,$(CHECKSUM)))

This runs nicely with a plain make but if I try make -j2 I get a segfault. I have captured this in gdb, output below

(gdb) bt
#0 eval_buffer (
buffer=0x819abb8 "CHECKSUM := ac2773ef9cb98feab91b3ddfeedc20019d1516c3")
at read.c:430
#1 0x08050012 in func_eval (o=0x819aae9 "[\230[", argv=0x0,
funcname=0x8062e38 "eval") at function.c:1369
#2 0x080506a8 in expand_builtin_function (o=0x819aae9 "[\230[",
argc=135900088, argv=0xfeffa8b0, entry_p=0xffffffca) at function.c:2115
#3 0x08050819 in handle_function (op=0xfeffa944, stringp=0x818ea38)
at function.c:2213
#4 0x0804c630 in variable_expand_string (line=0x0,
string=0x818ea00 "\t$(eval CHECKSUM := $(word 1,$(shell cat $^ | sha1sum))) $(if $(wildcard $(CACHEDIR)/$(CHECKSUM)), $(call do-copy,$(CHECKSUM)), $(call do-dct4-selection,$(CHECKSUM)))", length=-1) at expand.c:253
#5 0x0804c9a6 in variable_expand (
line=0x818ea00 "\t$(eval CHECKSUM := $(word 1,$(shell cat $^ | sha1sum))) $(if $(wildcard $(CACHEDIR)/$(CHECKSUM)), $(call do-copy,$(CHECKSUM)), $(call do-dct4-selection,$(CHECKSUM)))") at expand.c:418
#6 0x0804ca40 in variable_expand_for_file (
line=0xffffffca <Address 0xffffffca out of bounds>, file=0x819abb8)
at expand.c:463
#7 0x0804cc15 in allocated_variable_expand_for_file (
line=0x818ea00 "\t$(eval CHECKSUM := $(word 1,$(shell cat $^ | sha1sum))) $(if $(wildcard $(CACHEDIR)/$(CHECKSUM)), $(call do-copy,$(CHECKSUM)), $(call do-dct4-selection,$(CHECKSUM)))", file=0x80725b8) at expand.c:548
#8 0x08053319 in new_job (file=0x80725b8) at job.c:1600
#9 0x0805c28a in update_file_1 (file=0x80725b8, depth=0) at remake.c:761
#10 0x0805b984 in update_file (file=0x80725b8, depth=0) at remake.c:307
#11 0x0805b63f in update_goal_chain (goals=0x80ba8d8) at remake.c:154
#12 0x080551a2 in main (argc=7, argv=0xfeffc454, envp=0xfeffc474)
at main.c:2198
(gdb) print reading_file
$1 = (const struct floc *) 0x0

line 430 at read.c:

ebuf.floc = *reading_file;

So we try to dereference a NULL pointer here. As stated this is make 3.81.

I hope this is of help, otherwise I'll try to provide more information.

Cheers,
Rasmus

Anonymous

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach File(s):
   
   
Comment:
   

Attached Files
file #13043:  20033.diff added by psmith (2KiB - text/x-patch)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -unavailable- added by rasmus (Posted a comment)
  • -unavailable- added by psmith (Posted a comment)
  • -unavailable- added by mdorey (Posted a comment)
  •  

    Do you think this task is very important?
    If so, you can click here to add your encouragement to it.
    This task has 0 encouragements so far.

    Only logged-in users can vote.

     

    Please enter the title of George Orwell's famous dystopian book (it's a date):

     

     

    Follow 7 latest changes.

    Date Changed By Updated Field Previous Value => Replaced By
    Wed 28 Jul 2010 06:18:31 AM UTCpsmithFixed Release4.0=>3.82
    Thu 14 Jun 2007 02:27:57 AM UTCpsmithAttached File-=>Added 20033.diff, #13043
      StatusNone=>Fixed
      Assigned toNone=>psmith
      Open/ClosedOpen=>Closed
      Fixed ReleaseNone=>4.0
    Sat 02 Jun 2007 08:21:38 AM UTCrasmusCarbon-Copy-=>Added rasmus

    Back to the top


    Powered by Savane 3.1-cleanup1