Can't build first git commit

Can't build first git commit

[Fabio Aiuto]

Hi to all,

I'm trying to build first commit of git made by Linus. I mean the one
named e83c5163316f89bfbde7d9ab23... (I think this is enough).
But at building stage i have the following error:

make all 
gcc -g -o update-cache update-cache.o read-cache.o -lssl
/usr/bin/ld: update-cache.o: undefined reference to symbol
'SHA1_Update@@OPENSSL_1_1_0'
//usr/lib/i386-linux-gnu/libcrypto.so.1.1: error adding symbols: DSO
missing from command line
collect2: error: ld returned 1 exit status
make: *** [update-cache] Errore 1
Makefile:16: set di istruzioni per l'obiettivo "update-cache" non
riuscito

I run a debian stretch on my machine. Could anyone help me?
Thanks in advance.
Fabio.

Re: Can't build first git commit

[Santiago Torres]

[-- Attachment #1: Type: text/plain, Size: 1156 bytes --]

This commit is about 14 years old:

    Date:   Thu Apr 7 15:13:13 2005 -0700

Unless you have a toolchain from around that time, I'd be very surprised
if things build. Notably, there you're having an issue with the symbols
that lssl is exposing (I suspect you're not even using the openssl 1.0.0
series anymore.

Thanks,
-Santiago.

On Mon, Mar 04, 2019 at 08:58:37PM +0100, Fabio Aiuto wrote:
> Hi to all,
> 
> I'm trying to build first commit of git made by Linus. I mean the one
> named e83c5163316f89bfbde7d9ab23... (I think this is enough).
> But at building stage i have the following error:
> 
> make all 
> gcc -g -o update-cache update-cache.o read-cache.o -lssl
> /usr/bin/ld: update-cache.o: undefined reference to symbol
> 'SHA1_Update@@OPENSSL_1_1_0'
> //usr/lib/i386-linux-gnu/libcrypto.so.1.1: error adding symbols: DSO
> missing from command line
> collect2: error: ld returned 1 exit status
> make: *** [update-cache] Errore 1
> Makefile:16: set di istruzioni per l'obiettivo "update-cache" non
> riuscito
> 
> I run a debian stretch on my machine. Could anyone help me?
> Thanks in advance.
> Fabio.
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: Can't build first git commit

[Fabio Aiuto]

What a pity, It would have been very useful for me, to debug around
that simple version, to understand how everithing works.
Thank you Santiago.
Fabio.
Il giorno lun, 04/03/2019 alle 15.10 -0500, Santiago Torres ha scritto:
> This commit is about 14 years old:
> 
>     Date:   Thu Apr 7 15:13:13 2005 -0700
> 
> Unless you have a toolchain from around that time, I'd be very
> surprised
> if things build. Notably, there you're having an issue with the
> symbols
> that lssl is exposing (I suspect you're not even using the openssl
> 1.0.0
> series anymore.
> 
> Thanks,
> -Santiago.
> 
> On Mon, Mar 04, 2019 at 08:58:37PM +0100, Fabio Aiuto wrote:
> > Hi to all,
> > 
> > I'm trying to build first commit of git made by Linus. I mean the
> > one
> > named e83c5163316f89bfbde7d9ab23... (I think this is enough).
> > But at building stage i have the following error:
> > 
> > make all 
> > gcc -g -o update-cache update-cache.o read-cache.o -lssl
> > /usr/bin/ld: update-cache.o: undefined reference to symbol
> > 'SHA1_Update@@OPENSSL_1_1_0'
> > //usr/lib/i386-linux-gnu/libcrypto.so.1.1: error adding symbols:
> > DSO
> > missing from command line
> > collect2: error: ld returned 1 exit status
> > make: *** [update-cache] Errore 1
> > Makefile:16: set di istruzioni per l'obiettivo "update-cache" non
> > riuscito
> > 
> > I run a debian stretch on my machine. Could anyone help me?
> > Thanks in advance.
> > Fabio.
> > 

Re: Can't build first git commit

[Santiago Torres]

[-- Attachment #1: Type: text/plain, Size: 1983 bytes --]

It happens. 

Something that I liked playing with to get myself a better handle of how
git works is using libgit2 and its python bindings. I find the codebase
a little bit smaller and its bindings to higher-level languages can help
you abstract the memory management and other low-level stuff that is
sometimes distracting.

Good luck!
-Santiago.


On Mon, Mar 04, 2019 at 09:19:52PM +0100, Fabio Aiuto wrote:
> What a pity, It would have been very useful for me, to debug around
> that simple version, to understand how everithing works.
> Thank you Santiago.
> Fabio.
> Il giorno lun, 04/03/2019 alle 15.10 -0500, Santiago Torres ha scritto:
> > This commit is about 14 years old:
> > 
> >     Date:   Thu Apr 7 15:13:13 2005 -0700
> > 
> > Unless you have a toolchain from around that time, I'd be very
> > surprised
> > if things build. Notably, there you're having an issue with the
> > symbols
> > that lssl is exposing (I suspect you're not even using the openssl
> > 1.0.0
> > series anymore.
> > 
> > Thanks,
> > -Santiago.
> > 
> > On Mon, Mar 04, 2019 at 08:58:37PM +0100, Fabio Aiuto wrote:
> > > Hi to all,
> > > 
> > > I'm trying to build first commit of git made by Linus. I mean the
> > > one
> > > named e83c5163316f89bfbde7d9ab23... (I think this is enough).
> > > But at building stage i have the following error:
> > > 
> > > make all 
> > > gcc -g -o update-cache update-cache.o read-cache.o -lssl
> > > /usr/bin/ld: update-cache.o: undefined reference to symbol
> > > 'SHA1_Update@@OPENSSL_1_1_0'
> > > //usr/lib/i386-linux-gnu/libcrypto.so.1.1: error adding symbols:
> > > DSO
> > > missing from command line
> > > collect2: error: ld returned 1 exit status
> > > make: *** [update-cache] Errore 1
> > > Makefile:16: set di istruzioni per l'obiettivo "update-cache" non
> > > riuscito
> > > 
> > > I run a debian stretch on my machine. Could anyone help me?
> > > Thanks in advance.
> > > Fabio.
> > > 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: Can't build first git commit

[Jeff King]

On Mon, Mar 04, 2019 at 08:58:37PM +0100, Fabio Aiuto wrote:

> I'm trying to build first commit of git made by Linus. I mean the one
> named e83c5163316f89bfbde7d9ab23... (I think this is enough).
> But at building stage i have the following error:
> 
> make all 
> gcc -g -o update-cache update-cache.o read-cache.o -lssl
> /usr/bin/ld: update-cache.o: undefined reference to symbol
> 'SHA1_Update@@OPENSSL_1_1_0'

The sha1 routines are in libcrypto. See 3be4b61aa4 (Link with -lcrypto
instead of -lssl when using openssl libraries., 2005-05-10). I also
needed -lz. See 9426167765 (Add "-lz" to link line to get in zlib.,
2005-04-08).

You can patch the Makefile, or just override it like:

  make LIBS='-lcrypto -lz'

which builds for me on current Debian unstable. I don't think you can
actually fetch with that old build, but I used periodically check that
Git v1.0 can fetch happily from GitHub. I haven't in a while, so let me
know if you try it and it doesn't work. ;)

-Peff

Re: Can't build first git commit

[asymptosis]

On Mon, Mar 04, 2019 at 09:19:52PM +0100, Fabio Aiuto wrote:
> What a pity, It would have been very useful for me, to debug around
> that simple version, to understand how everithing works.

To understand how Git works, it is a good idea to work through the "Git
Internals" chapter in the Git Book:

https://git-scm.com/book/en/v2/Git-Internals-Plumbing-and-Porcelain

To help me understand that chapter, I started re-implementing Git as simple
bash scripts, which have been placed at https://notabug.org/cryptarch/sgit

Re: Can't build first git commit

[Jonathan Nieder]

Hi,

Fabio Aiuto wrote:

> What a pity, It would have been very useful for me, to debug around
> that simple version, to understand how everithing works.

Jeff King suggested how you can update the build command to get it
working.  In general, I think people sometimes overthink what is
involved in a build: something like

	cc *.c

plus appropriate warning, optimization, debug, -D, and library flags
should work just fine.

If you haven't already read it, I also recommend reading
https://www.kernel.org/pub/software/scm/git/docs/user-manual.html#hacking-git.
If you have ideas for improving the text there (perhaps with suggested
build instructions?), patches to Documentation/user-manual.txt are
very welcome.

Thanks and happy hacking,
Jonathan

Re: Can't build first git commit

[Fabio Aiuto]

Thank you Jeff I'll have news within few days!!!
Il giorno lun, 04/03/2019 alle 15.40 -0500, Jeff King ha scritto:
> On Mon, Mar 04, 2019 at 08:58:37PM +0100, Fabio Aiuto wrote:
> 
> > I'm trying to build first commit of git made by Linus. I mean the
> > one
> > named e83c5163316f89bfbde7d9ab23... (I think this is enough).
> > But at building stage i have the following error:
> > 
> > make all 
> > gcc -g -o update-cache update-cache.o read-cache.o -lssl
> > /usr/bin/ld: update-cache.o: undefined reference to symbol
> > 'SHA1_Update@@OPENSSL_1_1_0'
> 
> The sha1 routines are in libcrypto. See 3be4b61aa4 (Link with
> -lcrypto
> instead of -lssl when using openssl libraries., 2005-05-10). I also
> needed -lz. See 9426167765 (Add "-lz" to link line to get in zlib.,
> 2005-04-08).
> 
> You can patch the Makefile, or just override it like:
> 
>   make LIBS='-lcrypto -lz'
> 
> which builds for me on current Debian unstable. I don't think you can
> actually fetch with that old build, but I used periodically check
> that
> Git v1.0 can fetch happily from GitHub. I haven't in a while, so let
> me
> know if you try it and it doesn't work. ;)
> 
> -Peff

Re: Can't build first git commit

[Jeff King]

On Mon, Mar 04, 2019 at 03:40:07PM -0500, Jeff King wrote:

> You can patch the Makefile, or just override it like:
> 
>   make LIBS='-lcrypto -lz'
> 
> which builds for me on current Debian unstable. I don't think you can
> actually fetch with that old build, but I used periodically check that
> Git v1.0 can fetch happily from GitHub. I haven't in a while, so let me
> know if you try it and it doesn't work. ;)

I just tried this, and it does indeed work. I had to build v1.0.0 with

  make NO_OPENSSL=Nope

I think the issue is that some old code embedded openssl's BIGNUM in a
struct, and later versions of openssl stopped publicly defining the
types.

I was able to clone git://github.com/git/git with the result, though of
course it chokes no the sha1collisiondetection submodule. You can still
use "git log", though, and checkout older commits.

-Peff

Re: Can't build first git commit

[Fabio Aiuto]

Il giorno mar, 05/03/2019 alle 14.15 -0500, Jeff King ha scritto:
> On Mon, Mar 04, 2019 at 03:40:07PM -0500, Jeff King wrote:
> 
> > You can patch the Makefile, or just override it like:
> > 
> >   make LIBS='-lcrypto -lz'
> > 
> > which builds for me on current Debian unstable. I don't think you
> > can
> > actually fetch with that old build, but I used periodically check
> > that
> > Git v1.0 can fetch happily from GitHub. I haven't in a while, so
> > let me
> > know if you try it and it doesn't work. ;)
> 
> I just tried this, and it does indeed work. I had to build v1.0.0
> with
> 
>   make NO_OPENSSL=Nope
> 
> I think the issue is that some old code embedded openssl's BIGNUM in
> a
> struct, and later versions of openssl stopped publicly defining the
> types.
> 
> I was able to clone git://github.com/git/git with the result, though
> of
> course it chokes no the sha1collisiondetection submodule. You can
> still
> use "git log", though, and checkout older commits.
> 
> -Peff

Hi Jeff,
I've just typed make make LIBS='-lcrypto -lz' from the shell inside the
 
directory containing the whole first commit and it seems to work. I
trace all through the code. Yes the fetch command wasn't written at
that time, right? I didn't understand why should be better to work with
the git code from github. There's something I misunderstood?
Thank you Jeff ;-)

Re: Can't build first git commit

[Jeff King]

On Wed, Mar 06, 2019 at 08:58:40PM +0100, Fabio Aiuto wrote:

> Yes the fetch command wasn't written at that time, right? I didn't
> understand why should be better to work with the git code from github.
> There's something I misunderstood?

I just mean that it is an interesting fact that modern Git and Git v1.0
can still interact seamlessly over the network. I.e., you could still
collaborate with somebody using an ancient version of Git (hopefully
nobody is using v1.0, but logically it extends to all of the
intermediate versions).

-Peff

Re: Can't build first git commit

[Fabio Aiuto]

Yes that's great.
Why somebody told me about a outdate toolchain when I started this
little thread about first git compiling? Sorry for my stupid question
but the you just resolved with that make override...
Maybe there's something I have to know?

Il giorno mer, 06/03/2019 alle 15.03 -0500, Jeff King ha scritto:
> On Wed, Mar 06, 2019 at 08:58:40PM +0100, Fabio Aiuto wrote:
> 
> > Yes the fetch command wasn't written at that time, right? I didn't
> > understand why should be better to work with the git code from
> > github.
> > There's something I misunderstood?
> 
> I just mean that it is an interesting fact that modern Git and Git
> v1.0
> can still interact seamlessly over the network. I.e., you could still
> collaborate with somebody using an ancient version of Git (hopefully
> nobody is using v1.0, but logically it extends to all of the
> intermediate versions).
> 
> -Peff

Re: Can't build first git commit

[Santiago Torres]

[-- Attachment #1: Type: text/plain, Size: 1367 bytes --]

That was proably my bad. 

I still find it somewhat astounding that it compiles with a modern
toolchain after 15+ years. Many projects fail to do so (it's an
understandably high bar to have).

I'm glad this is possible and you were able to do so (maybe you want to
share your experiences about it somewhere? :))

Cheers!
-Santiago.


On Wed, Mar 06, 2019 at 09:19:58PM +0100, Fabio Aiuto wrote:
> Yes that's great.
> Why somebody told me about a outdate toolchain when I started this
> little thread about first git compiling? Sorry for my stupid question
> but the you just resolved with that make override...
> Maybe there's something I have to know?
> 
> Il giorno mer, 06/03/2019 alle 15.03 -0500, Jeff King ha scritto:
> > On Wed, Mar 06, 2019 at 08:58:40PM +0100, Fabio Aiuto wrote:
> > 
> > > Yes the fetch command wasn't written at that time, right? I didn't
> > > understand why should be better to work with the git code from
> > > github.
> > > There's something I misunderstood?
> > 
> > I just mean that it is an interesting fact that modern Git and Git
> > v1.0
> > can still interact seamlessly over the network. I.e., you could still
> > collaborate with somebody using an ancient version of Git (hopefully
> > nobody is using v1.0, but logically it extends to all of the
> > intermediate versions).
> > 
> > -Peff

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: Can't build first git commit

[Fabio Aiuto]

Found this error

Program received signal SIGSEGV, Segmentation fault.
0x00401628 in verify_hdr (hdr=0x0, size=32) at read-cache.c:192
192		if (hdr->signature != CACHE_SIGNATURE)

when tracing on line

static int verify_hdr(struct cache_header *hdr, unsigned long size)
{
	SHA_CTX c;
	unsigned char sha1[20];

	if (hdr->signature != CACHE_SIGNATURE)<--- this line

inside file read-cache.c
What's going on?

Il giorno mer, 06/03/2019 alle 15.03 -0500, Jeff King ha scritto:
> On Wed, Mar 06, 2019 at 08:58:40PM +0100, Fabio Aiuto wrote:
> 
> > Yes the fetch command wasn't written at that time, right? I didn't
> > understand why should be better to work with the git code from
> > github.
> > There's something I misunderstood?
> 
> I just mean that it is an interesting fact that modern Git and Git
> v1.0
> can still interact seamlessly over the network. I.e., you could still
> collaborate with somebody using an ancient version of Git (hopefully
> nobody is using v1.0, but logically it extends to all of the
> intermediate versions).
> 
> -Peff

Re: Can't build first git commit

[Jeff King]

On Wed, Mar 06, 2019 at 03:39:11PM -0500, Santiago Torres wrote:

> I still find it somewhat astounding that it compiles with a modern
> toolchain after 15+ years. Many projects fail to do so (it's an
> understandably high bar to have).

I think the key thing is that Git has very few external dependencies.
You'll note that I had to turn off OpenSSL support to get it to compile.
The only other library from back then is libz, which is extremely
stable.

Other than, it just depends on a reasonable C compiler and a POSIX libc,
both of which have been standardized for decades.

I suspect that today's Git will also compile pretty well in 15 years,
but you'll probably to say NO_GETTEXT and turn a few other knobs.
Libcurl might eventually be a problem, though I've been pretty impressed
with its stability and backwards-compatibility so far (and of course
it's optional if disabling http support is OK).

-Peff

Re: Can't build first git commit

[Jeff King]

On Wed, Mar 06, 2019 at 09:52:59PM +0100, Fabio Aiuto wrote:

> Found this error
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x00401628 in verify_hdr (hdr=0x0, size=32) at read-cache.c:192
> 192		if (hdr->signature != CACHE_SIGNATURE)
> 
> when tracing on line
> 
> static int verify_hdr(struct cache_header *hdr, unsigned long size)
> {
> 	SHA_CTX c;
> 	unsigned char sha1[20];
> 
> 	if (hdr->signature != CACHE_SIGNATURE)<--- this line
> 
> inside file read-cache.c
> What's going on?

Well, hdr is NULL, so you can't dereference it. Without knowing more
about how you called it, it's hard to say. But do note that the index
format has been updated once or twice (as well as the pack format). You
probably can't just chdir into a clone made by a recent Git and expect
ancient versions to read it.

You'll have to start a repo from scratch, or clone an existing with the
old version (which will be tricky; you might have to use a modern
upload-pack to represent the server side, but I think that first version
doesn't even have any fetch capabilities at all).

It might also be possible to convince modern Git to write out in old
formats that ancient Git will understand, but I don't know the right
config knobs to tweak offhand.

-Peff

Re: Can't build first git commit

[Jeff King]

[+cc git@vger; let's keep this on the list; I'm not sure there's a lot
 to be learned from fiddling with this old version, but if we're going
 to do it, let's at least record our attempts for posterity]

On Thu, Mar 07, 2019 at 08:04:55PM +0100, Fabio Aiuto wrote:

> I made a directory "trial" and inside it I ran init-db (that's the
> grand-father of git init) inside that directory. Then I created a file 
> this way:
> 
> echo 'helloooo!' >file.txt
> 
> then then executed update-cache file.txt (thinking that's the ancestor
> of git add), but that doesn't work and it returns with a segmentation
> fault, due to the nullity of the header. But in this first commit how
> do I add a file in the cache? What's the right way to build up a
> repository from scratch with commit e83c5163316f?

Hmm. I tried that, too, and got a segfault. Then I tried it again a few
minutes later, and it worked.

One thing that seems to provoke it consistently is having a zero-length
index file (because that causes mmap to return NULL, and the code does
not check for that case). Try this:

  # This works.
  init-db
  echo content >file
  update-cache file

  # This segfaults
  >.dircache/index
  update-cache file

  # And this works again
  rm .dircache/index
  update-cache file

I'm not sure why it would ever fail without that explicit empty-file
write. Perhaps there's some code path that writes out an empty index
file, and we inadvertently triggered it.

If you're going to play with such an antique commit, you have to expect
that things won't necessarily work, and be ready to poke around in the
debugger.

-Peff

Re: Can't build first git commit

[Fabio Aiuto]

Sorry for forgetting the cc!!!
Thank you so much for help, hope not to bore you all ;)
I think that I will go on studying git this way, and follow all the 
improvements that were made along his history. I think that
asymptosis too gave me a good link, related to that bash script
implmentation.
Thank you Jeff, now I'm trying to understand how could I show the
contents of the index in that version.

Fabio.

Il giorno gio, 07/03/2019 alle 14.41 -0500, Jeff King ha scritto:
> [+cc git@vger; let's keep this on the list; I'm not sure there's a
> lot
>  to be learned from fiddling with this old version, but if we're
> going
>  to do it, let's at least record our attempts for posterity]
> 
> On Thu, Mar 07, 2019 at 08:04:55PM +0100, Fabio Aiuto wrote:
> 
> > I made a directory "trial" and inside it I ran init-db (that's the
> > grand-father of git init) inside that directory. Then I created a
> > file 
> > this way:
> > 
> > echo 'helloooo!' >file.txt
> > 
> > then then executed update-cache file.txt (thinking that's the
> > ancestor
> > of git add), but that doesn't work and it returns with a
> > segmentation
> > fault, due to the nullity of the header. But in this first commit
> > how
> > do I add a file in the cache? What's the right way to build up a
> > repository from scratch with commit e83c5163316f?
> 
> Hmm. I tried that, too, and got a segfault. Then I tried it again a
> few
> minutes later, and it worked.
> 
> One thing that seems to provoke it consistently is having a zero-
> length
> index file (because that causes mmap to return NULL, and the code
> does
> not check for that case). Try this:
> 
>   # This works.
>   init-db
>   echo content >file
>   update-cache file
> 
>   # This segfaults
>   >.dircache/index
>   update-cache file
> 
>   # And this works again
>   rm .dircache/index
>   update-cache file
> 
> I'm not sure why it would ever fail without that explicit empty-file
> write. Perhaps there's some code path that writes out an empty index
> file, and we inadvertently triggered it.
> 
> If you're going to play with such an antique commit, you have to
> expect
> that things won't necessarily work, and be ready to poke around in
> the
> debugger.
> 
> -Peff