• Steven Schveighoffer (14/14) May 07 2016 I went to create a PR, and did a git pull on all my repositories. After
• Walter Bright (3/16) May 08 2016 __dmd_personality_v0 is in druntime\src\rt\dwarfeh.d, and was added rath...
• Steven Schveighoffer (35/58) May 08 2016 I don't know what I'm doing wrong:
• Jacob Carlborg (6/9) May 08 2016 I'm not sure what how your current setup looks like. But on OS X
• Walter Bright (4/24) May 08 2016 I should read more carefully, it is __Unwind_GetIPInfo that is undefined...
• Jacob Carlborg (7/11) May 08 2016 It's a bug that has been reported and fixed [1]. Unwind_GetIPInfo
• Steven Schveighoffer (4/16) May 08 2016 Ok thanks. Will wait for next release. For some reason this
• Jacob Carlborg (6/8) May 09 2016 Since you're using DVM, :), you can easily roll back to 2.070.0 as a
• Steven Schveighoffer (8/14) May 10 2016 Oh yes, that's what I did ;) I still have some nagging issues with dvm
• Jacob Carlborg (6/10) May 10 2016 I think there was some requirement that LDC should be able to build DMD....
• Lionello Lunesu (4/18) May 09 2016 I ran into this as well. It's a bug in the package from brew: it shipped...
• John Colvin (2/7) May 09 2016 In what way do you mean "shipped with the wrong phobos"?
• Lionello Lunesu (14/23) May 09 2016 Well, it's either wrong phobos or wrong dmd. I assumed phobos was
• Jacob Carlborg (7/10) May 09 2016 The issue Steven is having is that the symbol doesn't exist in the
• Lionello Lunesu (5/13) May 11 2016 Yeah, I see now. I thought it was because of the version "mismatch".
• John Colvin (5/19) May 10 2016 That's because for some reason the VERSION file is wrong in the
• Lionello Lunesu (2/21) May 11 2016 Ah, I see. Thanks!

Steven Schveighoffer <schveiguy yahoo.com> writes:

I went to create a PR, and did a git pull on all my repositories. After 
make clean on OSX, dmd would not build:

/Applications/Xcode.app/Contents/Developer/usr/bin/make -C src -f posix.mak
no cpu specified, assuming X86
CC=c++ dmd idgen.d
Undefined symbols for architecture x86_64:
   "__Unwind_GetIPInfo", referenced from:
       ___dmd_personality_v0 in libphobos2.a(dwarfeh_5c8_811.o)
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see 
invocation)

I had to revert to dmd 2.070 as the build tool to get it to build.

My question is: is this expected behavior, or should I file a bug?

-Steve

Walter Bright <newshound2 digitalmars.com> writes:

On 5/7/2016 11:43 PM, Steven Schveighoffer wrote:
 I went to create a PR, and did a git pull on all my repositories. After make
 clean on OSX, dmd would not build:

 /Applications/Xcode.app/Contents/Developer/usr/bin/make -C src -f posix.mak
 no cpu specified, assuming X86
 CC=c++ dmd idgen.d
 Undefined symbols for architecture x86_64:
   "__Unwind_GetIPInfo", referenced from:
       ___dmd_personality_v0 in libphobos2.a(dwarfeh_5c8_811.o)
 ld: symbol(s) not found for architecture x86_64
 clang: error: linker command failed with exit code 1 (use -v to see invocation)

 I had to revert to dmd 2.070 as the build tool to get it to build.

 My question is: is this expected behavior, or should I file a bug?

 -Steve

__dmd_personality_v0 is in druntime\src\rt\dwarfeh.d, and was added rather 
recently. So it seems you have a dmd-phobos mismatch.

Steven Schveighoffer <schveiguy yahoo.com> writes:

On 5/8/16 9:18 AM, Walter Bright wrote:
 On 5/7/2016 11:43 PM, Steven Schveighoffer wrote:
 I went to create a PR, and did a git pull on all my repositories.
 After make
 clean on OSX, dmd would not build:

 /Applications/Xcode.app/Contents/Developer/usr/bin/make -C src -f
 posix.mak
 no cpu specified, assuming X86
 CC=c++ dmd idgen.d
 Undefined symbols for architecture x86_64:
   "__Unwind_GetIPInfo", referenced from:
       ___dmd_personality_v0 in libphobos2.a(dwarfeh_5c8_811.o)
 ld: symbol(s) not found for architecture x86_64
 clang: error: linker command failed with exit code 1 (use -v to see
 invocation)

 I had to revert to dmd 2.070 as the build tool to get it to build.

 My question is: is this expected behavior, or should I file a bug?

 -Steve

 __dmd_personality_v0 is in druntime\src\rt\dwarfeh.d, and was added
 rather recently. So it seems you have a dmd-phobos mismatch.

I don't know what I'm doing wrong:

Stevens-MacBook-Pro:dmd steves$ dmd --version
DMD64 D Compiler v2.071.0
Copyright (c) 1999-2015 by Digital Mars written by Walter Bright
Stevens-MacBook-Pro:dmd steves$ type dmd
dmd is hashed (/Users/steves/.dvm/compilers/dmd-2.071.0/osx/bin/dmd)

I'm using dvm to install dmd, so I'm very sure it should work, I don't 
know if I have strange files somewhere?

If I build idgen manually and continue building, it fails again, this 
time with c++ compiler:

Stevens-MacBook-Pro:dmd steves$ cd src
Stevens-MacBook-Pro:src steves$ dmd idgen.d
Stevens-MacBook-Pro:src steves$ cd ..
Stevens-MacBook-Pro:dmd steves$ make -f posix.mak
/Applications/Xcode.app/Contents/Developer/usr/bin/make -C src -f posix.mak
no cpu specified, assuming X86
   (CC)  ROOT_OBJS  root/newdelete.c
c++ -c -Wno-deprecated -Wstrict-aliasing -Wno-logical-op-parentheses 
-Wno-dynamic-class-memaccess -Wno-switch -fno-exceptions -fno-rtti 
-D__pascal= -DMARS=1 -DTARGET_OSX=1 -DDM_TARGET_CPU_X86=1 -m64 -Iroot 
-MMD -MF newdelete.deps root/newdelete.c
   (CC)  GLUE_OBJS  glue.c
c++ -c -Wno-deprecated -Wstrict-aliasing -Wno-logical-op-parentheses 
-Wno-dynamic-class-memaccess -Wno-switch -fno-exceptions -fno-rtti 
-D__pascal= -DMARS=1 -DTARGET_OSX=1 -DDM_TARGET_CPU_X86=1 -m64 -Iroot 
-Itk -Ibackend -MMD -MF glue.deps glue.c
glue.c:26:10: fatal error: 'id.h' file not found
#include "id.h"
          ^
1 error generated.
make[1]: *** [glue.o] Error 1
make: *** [all] Error 2

Does this give any more clues?

-Steve

Jacob Carlborg <doob me.com> writes:

On Sunday, 8 May 2016 at 07:18:05 UTC, Walter Bright wrote:

 __dmd_personality_v0 is in druntime\src\rt\dwarfeh.d, and was 
 added rather recently. So it seems you have a dmd-phobos 
 mismatch.

I'm not sure what how your current setup looks like. But on OS X 
these days we should target clang, libc++ and libc++abi. 
Unwind_GetIPInfo does not exist in libc++abi.

--
/Jacob Carlborg

Walter Bright <newshound2 digitalmars.com> writes:

On 5/8/2016 12:18 AM, Walter Bright wrote:
 On 5/7/2016 11:43 PM, Steven Schveighoffer wrote:
 I went to create a PR, and did a git pull on all my repositories. After make
 clean on OSX, dmd would not build:

 /Applications/Xcode.app/Contents/Developer/usr/bin/make -C src -f posix.mak
 no cpu specified, assuming X86
 CC=c++ dmd idgen.d
 Undefined symbols for architecture x86_64:
   "__Unwind_GetIPInfo", referenced from:
       ___dmd_personality_v0 in libphobos2.a(dwarfeh_5c8_811.o)
 ld: symbol(s) not found for architecture x86_64
 clang: error: linker command failed with exit code 1 (use -v to see invocation)

 I had to revert to dmd 2.070 as the build tool to get it to build.

 My question is: is this expected behavior, or should I file a bug?

 -Steve

 __dmd_personality_v0 is in druntime\src\rt\dwarfeh.d, and was added rather
 recently. So it seems you have a dmd-phobos mismatch.

I should read more carefully, it is __Unwind_GetIPInfo that is undefined, not 
__dmd_personality_v0.

Anyhow, thanks to Jacob for correctly diagnosing it.

Jacob Carlborg <doob me.com> writes:

On Sunday, 8 May 2016 at 06:43:38 UTC, Steven Schveighoffer wrote:

 Undefined symbols for architecture x86_64:
   "__Unwind_GetIPInfo", referenced from:
       ___dmd_personality_v0 in libphobos2.a(dwarfeh_5c8_811.o)
 ld: symbol(s) not found for architecture x86_64

It's a bug that has been reported and fixed [1]. Unwind_GetIPInfo 
is not available on OS X or rather in libc++abi which is used by 
default on OS X.

[1] https://issues.dlang.org/show_bug.cgi?id=15911
--
/Jacob Carlborg

Steven Schveighoffer <schveiguy yahoo.com> writes:

On Sunday, 8 May 2016 at 10:30:06 UTC, Jacob Carlborg wrote:
 On Sunday, 8 May 2016 at 06:43:38 UTC, Steven Schveighoffer 
 wrote:

 Undefined symbols for architecture x86_64:
   "__Unwind_GetIPInfo", referenced from:
       ___dmd_personality_v0 in libphobos2.a(dwarfeh_5c8_811.o)
 ld: symbol(s) not found for architecture x86_64

 It's a bug that has been reported and fixed [1]. 
 Unwind_GetIPInfo is not available on OS X or rather in 
 libc++abi which is used by default on OS X.

 [1] https://issues.dlang.org/show_bug.cgi?id=15911
 --
 /Jacob Carlborg

Ok thanks. Will wait for next release. For some reason this 
doesn't affect command line compiles

-Steve

Jacob Carlborg <doob me.com> writes:

On 2016-05-08 12:33, Steven Schveighoffer wrote:

 Ok thanks. Will wait for next release. For some reason this doesn't
 affect command line compiles

Since you're using DVM, :), you can easily roll back to 2.070.0 as a 
workaround.

If I recall correctly, DMD is buildable with DMD 2.067.0 and later.

-- 
/Jacob Carlborg

Steven Schveighoffer <schveiguy yahoo.com> writes:

On 5/10/16 2:41 AM, Jacob Carlborg wrote:
 On 2016-05-08 12:33, Steven Schveighoffer wrote:

 Ok thanks. Will wait for next release. For some reason this doesn't
 affect command line compiles

 Since you're using DVM, :), you can easily roll back to 2.070.0 as a
 workaround.

Oh yes, that's what I did ;) I still have some nagging issues with dvm 
(I've reported them already), but otherwise, fantastic tool.

 If I recall correctly, DMD is buildable with DMD 2.067.0 and later.

I couldn't remember the exact version, but since 2.070.2 worked, I 
didn't look any further. I just was curious whether the policy is that 
this is an expectation that you have to use a prior version to build 
dmd, or if we should report bugs when dmd can't build itself.

-Steve

Jacob Carlborg <doob me.com> writes:

On 2016-05-10 14:29, Steven Schveighoffer wrote:

 I couldn't remember the exact version, but since 2.070.2 worked, I
 didn't look any further. I just was curious whether the policy is that
 this is an expectation that you have to use a prior version to build
 dmd, or if we should report bugs when dmd can't build itself.

I think there was some requirement that LDC should be able to build DMD. 
So they just picked the current version of LDC at that time (when 
switching to D) and made sure it worked.

-- 
/Jacob Carlborg

Lionello Lunesu <lionello lunesu.remove.com> writes:

On 8/5/2016 14:43, Steven Schveighoffer wrote:
 I went to create a PR, and did a git pull on all my repositories. After
 make clean on OSX, dmd would not build:

 /Applications/Xcode.app/Contents/Developer/usr/bin/make -C src -f posix.mak
 no cpu specified, assuming X86
 CC=c++ dmd idgen.d
 Undefined symbols for architecture x86_64:
    "__Unwind_GetIPInfo", referenced from:
        ___dmd_personality_v0 in libphobos2.a(dwarfeh_5c8_811.o)
 ld: symbol(s) not found for architecture x86_64
 clang: error: linker command failed with exit code 1 (use -v to see
 invocation)

 I had to revert to dmd 2.070 as the build tool to get it to build.

 My question is: is this expected behavior, or should I file a bug?

 -Steve

I ran into this as well. It's a bug in the package from brew: it shipped 
with the wrong phobos. You can build your own DMD:

$ make -f posix.mak AUTO_BOOTSTRAP=1

John Colvin <john.loughran.colvin gmail.com> writes:

On Monday, 9 May 2016 at 07:57:33 UTC, Lionello Lunesu wrote:
 On 8/5/2016 14:43, Steven Schveighoffer wrote:
 [...]

 I ran into this as well. It's a bug in the package from brew: 
 it shipped with the wrong phobos. You can build your own DMD:

 $ make -f posix.mak AUTO_BOOTSTRAP=1

In what way do you mean "shipped with the wrong phobos"?

Lionello Lunesu <lionello lunesu.remove.com> writes:

On 9/5/2016 20:19, John Colvin wrote:
 On Monday, 9 May 2016 at 07:57:33 UTC, Lionello Lunesu wrote:
 On 8/5/2016 14:43, Steven Schveighoffer wrote:
 [...]

 I ran into this as well. It's a bug in the package from brew: it
 shipped with the wrong phobos. You can build your own DMD:

 $ make -f posix.mak AUTO_BOOTSTRAP=1

 In what way do you mean "shipped with the wrong phobos"?

Well, it's either wrong phobos or wrong dmd. I assumed phobos was 
mismatched, but perhaps DMD was mismatched? Installing with brew suggest 
it's installing 2.071.0, but DMD appears to be 2.070:

$ brew install dmd
==> Downloading 
https://homebrew.bintray.com/bottles/dmd-2.071.0.el_capitan.bottle.tar.gz
Already downloaded: 
/Library/Caches/Homebrew/dmd-2.071.0.el_capitan.bottle.tar.gz
==> Pouring dmd-2.071.0.el_capitan.bottle.tar.gz
🍺  /usr/local/Cellar/dmd/2.071.0: 561 files, 65.0M

$ /usr/local/bin/dmd --version
DMD64 D Compiler v2.070
Copyright (c) 1999-2015 by Digital Mars written by Walter Bright

Jacob Carlborg <doob me.com> writes:

On 2016-05-10 06:49, Lionello Lunesu wrote:

 Well, it's either wrong phobos or wrong dmd. I assumed phobos was
 mismatched, but perhaps DMD was mismatched? Installing with brew suggest
 it's installing 2.071.0, but DMD appears to be 2.070:

The issue Steven is having is that the symbol doesn't exist in the 
default C++ runtime. This has been fixed in master. He installed DMD 
using DVM so it's not a Brew packaging problem.

Rolling back to 2.70.0 can be used as a workaround.

-- 
/Jacob Carlborg

Lionello Lunesu <lionello lunesu.remove.com> writes:

On 10/5/2016 14:39, Jacob Carlborg wrote:
 On 2016-05-10 06:49, Lionello Lunesu wrote:

 Well, it's either wrong phobos or wrong dmd. I assumed phobos was
 mismatched, but perhaps DMD was mismatched? Installing with brew suggest
 it's installing 2.071.0, but DMD appears to be 2.070:

 The issue Steven is having is that the symbol doesn't exist in the
 default C++ runtime. This has been fixed in master. He installed DMD
 using DVM so it's not a Brew packaging problem.

 Rolling back to 2.70.0 can be used as a workaround.

Yeah, I see now. I thought it was because of the version "mismatch". 
Didn't know it was an actual bug. I had build my own and it worked. I 
guess I had gotten the fix.

Thanks for correcting me.

John Colvin <john.loughran.colvin gmail.com> writes:

On Tuesday, 10 May 2016 at 04:49:33 UTC, Lionello Lunesu wrote:
 Well, it's either wrong phobos or wrong dmd. I assumed phobos 
 was mismatched, but perhaps DMD was mismatched? Installing with 
 brew suggest it's installing 2.071.0, but DMD appears to be 
 2.070:

 $ brew install dmd
 ==> Downloading 
 https://homebrew.bintray.com/bottles/dmd-2.071.0.el_capitan.bottle.tar.gz
 Already downloaded: 
 /Library/Caches/Homebrew/dmd-2.071.0.el_capitan.bottle.tar.gz
 ==> Pouring dmd-2.071.0.el_capitan.bottle.tar.gz
 🍺  /usr/local/Cellar/dmd/2.071.0: 561 files, 65.0M

 $ /usr/local/bin/dmd --version
 DMD64 D Compiler v2.070
 Copyright (c) 1999-2015 by Digital Mars written by Walter Bright

That's because for some reason the VERSION file is wrong in the 
dmd github release.

Should be fixed by 
https://github.com/Homebrew/homebrew-core/pull/1040

Lionello Lunesu <lionello lunesu.remove.com> writes:

On 10/5/2016 16:24, John Colvin wrote:
 On Tuesday, 10 May 2016 at 04:49:33 UTC, Lionello Lunesu wrote:
 Well, it's either wrong phobos or wrong dmd. I assumed phobos was
 mismatched, but perhaps DMD was mismatched? Installing with brew
 suggest it's installing 2.071.0, but DMD appears to be 2.070:

 $ brew install dmd
 ==> Downloading
 https://homebrew.bintray.com/bottles/dmd-2.071.0.el_capitan.bottle.tar.gz
 Already downloaded:
 /Library/Caches/Homebrew/dmd-2.071.0.el_capitan.bottle.tar.gz
 ==> Pouring dmd-2.071.0.el_capitan.bottle.tar.gz
 🍺  /usr/local/Cellar/dmd/2.071.0: 561 files, 65.0M

 $ /usr/local/bin/dmd --version
 DMD64 D Compiler v2.070
 Copyright (c) 1999-2015 by Digital Mars written by Walter Bright

 That's because for some reason the VERSION file is wrong in the dmd
 github release.

 Should be fixed by https://github.com/Homebrew/homebrew-core/pull/1040

Ah, I see. Thanks!