From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
Hello,
The first patch is not that big a deal, but it's better if we're
freeing the refspecs, we might as well free all of them.
The second patch teaches get_stale_heads to use the user-provided
refspecs instead of the ones in the config. For example, running
git fetch --prune origin refs/heads/master:refs/heads/master
doesn't remove the other branches anymore. For a more interesting (and
believable) example, let's take
git fetch --prune origin refs/heads/b/*:refs/heads/b/*
because you want to prune the refs inside the b/ namespace
only. Currently git will delete all the refs that aren't under that
namespace. With the second patch applied, git won't remove any refs
outside the b/ namespace.
What is probably the most usual case is covered by the third patch,
which pretends that a "refs/tags/*:refs/tags/*" refspec was given on
the command-line.
Cheers,
cmn
Carlos Martín Nieto (3):
fetch: free all the additional refspecs
fetch: honor the user-provided refspecs when pruning refs
fetch: treat --tags like refs/tags/*:refs/tags/* when pruning
builtin/fetch.c | 19 ++++++++++---
builtin/remote.c | 2 +-
remote.c | 74 ++++++++++++++++++++++++++++++++++++++++++++++++-----
remote.h | 3 +-
4 files changed, 84 insertions(+), 14 deletions(-)
--
1.7.5.2.354.g349bf
From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
If the user gave us refspecs on the command line, we should use those
when deciding whether to prune a ref instead of relying on the
refspecs in the config.
Previously, running
git fetch --prune origin refs/heads/master:refs/remotes/origin/master
would delete every other tag under the origin namespace because we
were using the refspec to filter the available refs but using the
configured refspec to figure out if a ref had been deleted on the
remote. This is clearly the wrong thing to do.
Teach get_stale_heads about user-provided refspecs and use them if
they're available.
Signed-off-by: Carlos Martín Nieto <redacted>
---
builtin/fetch.c | 6 ++--
builtin/remote.c | 2 +-
remote.c | 74 ++++++++++++++++++++++++++++++++++++++++++++++++-----
remote.h | 3 +-
4 files changed, 73 insertions(+), 12 deletions(-)
@@ -505,10 +505,10 @@ static int fetch_refs(struct transport *transport, struct ref *ref_map)returnret;}-staticintprune_refs(structtransport*transport,structref*ref_map)+staticintprune_refs(structtransport*transport,structrefspec*refs,intn,structref*ref_map){intresult=0;-structref*ref,*stale_refs=get_stale_heads(transport->remote,ref_map);+structref*ref,*stale_refs=get_stale_heads(transport->remote,ref_map,refs,n);constchar*dangling_msg=dry_run?_(" (%s will become dangling)\n"):_(" (%s has become dangling)\n");
@@ -700,7 +700,7 @@ static int do_fetch(struct transport *transport,return1;}if(prune)-prune_refs(transport,ref_map);+prune_refs(transport,refs,ref_count,ref_map);free_refs(ref_map);/* if neither --no-tags nor --tags was specified, do automated tag
From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
If --tags is specified, add that refspec to the list given to prune_refs
so it knows to treat it as a filter on what refs to should consider
for prunning. This way
git fetch --prune --tags origin
only prunes tags and doesn't delete the branch refs.
Signed-off-by: Carlos Martín Nieto <redacted>
---
builtin/fetch.c | 11 ++++++++++-
1 files changed, 10 insertions(+), 1 deletions(-)
@@ -699,8 +699,17 @@ static int do_fetch(struct transport *transport,free_refs(ref_map);return1;}-if(prune)+if(prune){+/* If --tags was specified, we need to tell prune_refs+*thatwe'refilteringtherefsfromtheremote*/+if(tags==TAGS_SET){+constchar*tags_refspec="refs/tags/*:refs/tags/*";+refs=xrealloc(refs,(ref_count+1)*sizeof(structrefspec));+refs[ref_count]=*parse_fetch_refspec(1,&tags_refspec);+ref_count++;+}prune_refs(transport,refs,ref_count,ref_map);+}free_refs(ref_map);/* if neither --no-tags nor --tags was specified, do automated tag
From: Ben Boeckel <hidden> Date: 2016-06-15 22:52:12
On Thu, Oct 06, 2011 at 22:19:42 +0200, Carlos Martín Nieto wrote:
The first patch is not that big a deal, but it's better if we're
freeing the refspecs, we might as well free all of them.
The second patch teaches get_stale_heads to use the user-provided
refspecs instead of the ones in the config. For example, running
git fetch --prune origin refs/heads/master:refs/heads/master
doesn't remove the other branches anymore. For a more interesting (and
believable) example, let's take
git fetch --prune origin refs/heads/b/*:refs/heads/b/*
because you want to prune the refs inside the b/ namespace
only. Currently git will delete all the refs that aren't under that
namespace. With the second patch applied, git won't remove any refs
outside the b/ namespace.
What is probably the most usual case is covered by the third patch,
which pretends that a "refs/tags/*:refs/tags/*" refspec was given on
the command-line.
I applied the patches to current master (7f41b6b) and got a segfault
with:
git fetch -p -t origin master
It does not happen with master.
Backtrace:
(gdb) bt
#0 0x00007ffff7395d18 in __strchr_sse42 () from /lib64/libc.so.6
#1 0x00000000004b2d39 in find_in_refs (query=0x7fffffffdb90, ref_count=2, refs=<optimized out>) at remote.c:1709
#2 get_stale_heads_cb (refname=0x7a8f31 "refs/heads/a/branch/name", sha1=0x7a8f09 "\367\343\375C٩\223u\305OG\233)z\347X\370\333\325", <incomplete sequence \335>, flags=0, cb_data=0x7fffffffdc50) at remote.c:1740
#3 0x00000000004adf19 in do_for_each_ref (submodule=<optimized out>, base=0x4ea1c2 "", fn=0x4b2ca0 <get_stale_heads_cb>, trim=0, flags=0, cb_data=0x7fffffffdc50) at refs.c:684
#4 0x00000000004b4249 in get_stale_heads (remote=<optimized out>, fetch_map=<optimized out>, refs=<optimized out>, ref_count=<optimized out>) at remote.c:1777
#5 0x0000000000426cfb in prune_refs (ref_map=<optimized out>, n=<optimized out>, refs=<optimized out>, transport=0x78e040) at builtin/fetch.c:511
#6 do_fetch (ref_count=<optimized out>, refs=<optimized out>, transport=0x78e040) at builtin/fetch.c:711
#7 fetch_one (remote=<optimized out>, argc=<optimized out>, argv=<optimized out>) at builtin/fetch.c:894
#8 0x0000000000427550 in cmd_fetch (argc=2, argv=0x7fffffffe070, prefix=0x0) at builtin/fetch.c:955
#9 0x0000000000405084 in run_builtin (argv=0x7fffffffe070, argc=5, p=0x731b08) at git.c:308
#10 handle_internal_command (argc=5, argv=0x7fffffffe070) at git.c:466
#11 0x000000000040448b in run_argv (argv=0x7fffffffdf10, argcp=0x7fffffffdf1c) at git.c:512
#12 main (argc=5, argv=0x7fffffffe070) at git.c:585
--Ben
From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
On Thu, 2011-10-06 at 16:51 -0400, Ben Boeckel wrote:
On Thu, Oct 06, 2011 at 22:19:42 +0200, Carlos Martín Nieto wrote:
quoted
The first patch is not that big a deal, but it's better if we're
freeing the refspecs, we might as well free all of them.
The second patch teaches get_stale_heads to use the user-provided
refspecs instead of the ones in the config. For example, running
git fetch --prune origin refs/heads/master:refs/heads/master
doesn't remove the other branches anymore. For a more interesting (and
believable) example, let's take
git fetch --prune origin refs/heads/b/*:refs/heads/b/*
because you want to prune the refs inside the b/ namespace
only. Currently git will delete all the refs that aren't under that
namespace. With the second patch applied, git won't remove any refs
outside the b/ namespace.
What is probably the most usual case is covered by the third patch,
which pretends that a "refs/tags/*:refs/tags/*" refspec was given on
the command-line.
I applied the patches to current master (7f41b6b) and got a segfault
with:
git fetch -p -t origin master
It does not happen with master.
I thought I'd got rid of those problems. Thanks for noticing. I'll
investigate.
Backtrace:
(gdb) bt
#0 0x00007ffff7395d18 in __strchr_sse42 () from /lib64/libc.so.6
#1 0x00000000004b2d39 in find_in_refs (query=0x7fffffffdb90, ref_count=2, refs=<optimized out>) at remote.c:1709
#2 get_stale_heads_cb (refname=0x7a8f31 "refs/heads/a/branch/name", sha1=0x7a8f09 "\367\343\375C٩\223u\305OG\233)z\347X\370\333\325", <incomplete sequence \335>, flags=0, cb_data=0x7fffffffdc50) at remote.c:1740
#3 0x00000000004adf19 in do_for_each_ref (submodule=<optimized out>, base=0x4ea1c2 "", fn=0x4b2ca0 <get_stale_heads_cb>, trim=0, flags=0, cb_data=0x7fffffffdc50) at refs.c:684
#4 0x00000000004b4249 in get_stale_heads (remote=<optimized out>, fetch_map=<optimized out>, refs=<optimized out>, ref_count=<optimized out>) at remote.c:1777
#5 0x0000000000426cfb in prune_refs (ref_map=<optimized out>, n=<optimized out>, refs=<optimized out>, transport=0x78e040) at builtin/fetch.c:511
#6 do_fetch (ref_count=<optimized out>, refs=<optimized out>, transport=0x78e040) at builtin/fetch.c:711
#7 fetch_one (remote=<optimized out>, argc=<optimized out>, argv=<optimized out>) at builtin/fetch.c:894
#8 0x0000000000427550 in cmd_fetch (argc=2, argv=0x7fffffffe070, prefix=0x0) at builtin/fetch.c:955
#9 0x0000000000405084 in run_builtin (argv=0x7fffffffe070, argc=5, p=0x731b08) at git.c:308
#10 handle_internal_command (argc=5, argv=0x7fffffffe070) at git.c:466
#11 0x000000000040448b in run_argv (argv=0x7fffffffdf10, argcp=0x7fffffffdf1c) at git.c:512
#12 main (argc=5, argv=0x7fffffffe070) at git.c:585
--Ben
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
Hello,
The first patch is not that big a deal, but it's better if we're
freeing the refspecs, we might as well free all of them.
The second patch teaches get_stale_heads to use the user-provided
refspecs instead of the ones in the config. For example, running
git fetch --prune origin refs/heads/master:refs/heads/master
doesn't remove the other branches anymore. For a more interesting (and
believable) example, let's take
git fetch --prune origin refs/heads/b/*:refs/heads/b/*
because you want to prune the refs inside the b/ namespace
only. Currently git will delete all the refs that aren't under that
namespace. With the second patch applied, git won't remove any refs
outside the b/ namespace.
What is probably the most usual case is covered by the third patch,
which pretends that a "refs/tags/*:refs/tags/*" refspec was given on
the command-line.
Version 1 assumed that a refspec would have its dst filled
automatically. This is not the case and was fixed in the second patch.
Cheers,
cmn
Carlos Martín Nieto (3):
fetch: free all the additional refspecs
fetch: honor the user-provided refspecs when pruning refs
fetch: treat --tags like refs/tags/*:refs/tags/* when pruning
builtin/fetch.c | 19 ++++++++++---
builtin/remote.c | 2 +-
remote.c | 74 ++++++++++++++++++++++++++++++++++++++++++++++++-----
remote.h | 3 +-
4 files changed, 84 insertions(+), 14 deletions(-)
--
1.7.5.2.354.g349bf
From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
If --tags is specified, add that refspec to the list given to prune_refs
so it knows to treat it as a filter on what refs to should consider
for prunning. This way
git fetch --prune --tags origin
only prunes tags and doesn't delete the branch refs.
Signed-off-by: Carlos Martín Nieto <redacted>
---
builtin/fetch.c | 11 ++++++++++-
1 files changed, 10 insertions(+), 1 deletions(-)
@@ -699,8 +699,17 @@ static int do_fetch(struct transport *transport,free_refs(ref_map);return1;}-if(prune)+if(prune){+/* If --tags was specified, we need to tell prune_refs+*thatwe'refilteringtherefsfromtheremote*/+if(tags==TAGS_SET){+constchar*tags_refspec="refs/tags/*:refs/tags/*";+refs=xrealloc(refs,(ref_count+1)*sizeof(structrefspec));+refs[ref_count]=*parse_fetch_refspec(1,&tags_refspec);+ref_count++;+}prune_refs(transport,refs,ref_count,ref_map);+}free_refs(ref_map);/* if neither --no-tags nor --tags was specified, do automated tag
From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
If the user gave us refspecs on the command line, we should use those
when deciding whether to prune a ref instead of relying on the
refspecs in the config.
Previously, running
git fetch --prune origin refs/heads/master:refs/remotes/origin/master
would delete every other tag under the origin namespace because we
were using the refspec to filter the available refs but using the
configured refspec to figure out if a ref had been deleted on the
remote. This is clearly the wrong thing to do.
Teach get_stale_heads about user-provided refspecs and use them if
they're available.
Signed-off-by: Carlos Martín Nieto <redacted>
---
builtin/fetch.c | 6 ++--
builtin/remote.c | 2 +-
remote.c | 78 +++++++++++++++++++++++++++++++++++++++++++++++++-----
remote.h | 3 +-
4 files changed, 77 insertions(+), 12 deletions(-)
@@ -505,10 +505,10 @@ static int fetch_refs(struct transport *transport, struct ref *ref_map)returnret;}-staticintprune_refs(structtransport*transport,structref*ref_map)+staticintprune_refs(structtransport*transport,structrefspec*refs,intn,structref*ref_map){intresult=0;-structref*ref,*stale_refs=get_stale_heads(transport->remote,ref_map);+structref*ref,*stale_refs=get_stale_heads(transport->remote,ref_map,refs,n);constchar*dangling_msg=dry_run?_(" (%s will become dangling)\n"):_(" (%s has become dangling)\n");
@@ -700,7 +700,7 @@ static int do_fetch(struct transport *transport,return1;}if(prune)-prune_refs(transport,ref_map);+prune_refs(transport,refs,ref_count,ref_map);free_refs(ref_map);/* if neither --no-tags nor --tags was specified, do automated tag
@@ -1684,26 +1684,88 @@ struct stale_heads_info {structremote*remote;structstring_list*ref_names;structref**stale_refs_tail;+structrefspec*refs;+intref_count;};+/*+*Findarefspectoaremote's+*Returns0onsuccess,-1ifitcouldn'tfindatherefspec+*/+staticintfind_in_refs(structrefspec*refs,intref_count,structrefspec*query)+{+inti;+structrefspec*refspec;++for(i=0;i<ref_count;++i){+refspec=&refs[i];++/* No dst means it can't be used for prunning. */+if(!refspec->dst)+continue;++/*+*No'*'meansthatitmustmatchexactly.Ifitdoes+*haveit,trytomatchitagainstthepattern.If+*therefspecmatches,storetherefnameasitwould+*appearintheserverinquery->src.+*/+if(!strchr(refspec->dst,'*')){+if(!strcmp(query->dst,refspec->dst)){+query->src=xstrdup(refspec->src);+return0;+}+}else{+if(match_name_with_pattern(refspec->dst,query->dst,+refspec->src,&query->src)){+return0;+}+}+}++return-1;+}+staticintget_stale_heads_cb(constchar*refname,constunsignedchar*sha1,intflags,void*cb_data){structstale_heads_info*info=cb_data;structrefspecrefspec;+intret;memset(&refspec,0,sizeof(refspec));refspec.dst=(char*)refname;-if(!remote_find_tracking(info->remote,&refspec)){-if(!((flags&REF_ISSYMREF)||-string_list_has_string(info->ref_names,refspec.src))){-structref*ref=make_linked_ref(refname,&info->stale_refs_tail);-hashcpy(ref->new_sha1,sha1);-}++/*+*Iftheuserspeicifiedrefspecsonthecommandline,we+*shouldonlyusethosetocheck.Otherwise,lookinthe+*remote'sconfigurationforthebranch.+*/+if(info->ref_count)+ret=find_in_refs(info->refs,info->ref_count,&refspec);+else+ret=remote_find_tracking(info->remote,&refspec);++/* No matches */+if(ret)+return0;++/*+*Ifwedidfindasuitablerefspecandit'snotasymrefand+*it'snotinthelistofrefsthatcurrentlyexistinthat+*remoteweconsiderittobestale.+*/+if(!((flags&REF_ISSYMREF)||+string_list_has_string(info->ref_names,refspec.src))){+structref*ref=make_linked_ref(refname,&info->stale_refs_tail);+hashcpy(ref->new_sha1,sha1);}++free(refspec.src);return0;}-structref*get_stale_heads(structremote*remote,structref*fetch_map)+structref*get_stale_heads(structremote*remote,structref*fetch_map,+structrefspec*refs,intref_count){structref*ref,*stale_refs=NULL;structstring_listref_names=STRING_LIST_INIT_NODUP;
From: Jeff King <hidden> Date: 2016-06-15 22:52:12
On Thu, Oct 06, 2011 at 11:21:46PM +0200, Carlos Martín Nieto wrote:
If the user gave us refspecs on the command line, we should use those
when deciding whether to prune a ref instead of relying on the
refspecs in the config.
Previously, running
git fetch --prune origin refs/heads/master:refs/remotes/origin/master
would delete every other tag under the origin namespace because we
static int get_stale_heads_cb(const char *refname,
const unsigned char *sha1, int flags, void *cb_data)
{
struct stale_heads_info *info = cb_data;
struct refspec refspec;
+ int ret;
memset(&refspec, 0, sizeof(refspec));
refspec.dst = (char *)refname;
- if (!remote_find_tracking(info->remote, &refspec)) {
- if (!((flags & REF_ISSYMREF) ||
- string_list_has_string(info->ref_names, refspec.src))) {
- struct ref *ref = make_linked_ref(refname, &info->stale_refs_tail);
- hashcpy(ref->new_sha1, sha1);
- }
+
+ /*
+ * If the user speicified refspecs on the command line, we
+ * should only use those to check. Otherwise, look in the
+ * remote's configuration for the branch.
+ */
+ if (info->ref_count)
+ ret = find_in_refs(info->refs, info->ref_count, &refspec);
+ else
+ ret = remote_find_tracking(info->remote, &refspec);
Minor typo in the comment. But more importantly, this feels like a very
low-level place to be thinking about things like what the user gave us
on the command line.
Shouldn't get_stale_heads not get a remote at all, and just get a set of
refspecs? Those should be the minimal information it needs to get its
answer, right?
-Peff
@@ -699,8 +699,17 @@ static int do_fetch(struct transport *transport,free_refs(ref_map);return1;}-if(prune)+if(prune){+/* If --tags was specified, we need to tell prune_refs+*thatwe'refilteringtherefsfromtheremote*/+if(tags==TAGS_SET){+constchar*tags_refspec="refs/tags/*:refs/tags/*";+refs=xrealloc(refs,(ref_count+1)*sizeof(structrefspec));+refs[ref_count]=*parse_fetch_refspec(1,&tags_refspec);+ref_count++;+}prune_refs(transport,refs,ref_count,ref_map);+}
I don't think we can realloc refs here. It's passed into do_fetch. When
we realloc it, the old pointer value will be invalid. But when we return
from do_fetch, the caller (fetch_one) will still have that old value,
and will call free() on it.
Instead, you have to make a whole new list, copy the old values in, add
your new one, and then free the result.
-Peff
From: Carlos Martín Nieto <hidden> Date: 2016-06-15 22:52:12
On Fri, 2011-10-07 at 12:26 -0400, Jeff King wrote:
On Thu, Oct 06, 2011 at 11:21:46PM +0200, Carlos Martín Nieto wrote:
quoted
If the user gave us refspecs on the command line, we should use those
when deciding whether to prune a ref instead of relying on the
refspecs in the config.
Previously, running
git fetch --prune origin refs/heads/master:refs/remotes/origin/master
would delete every other tag under the origin namespace because we
static int get_stale_heads_cb(const char *refname,
const unsigned char *sha1, int flags, void *cb_data)
{
struct stale_heads_info *info = cb_data;
struct refspec refspec;
+ int ret;
memset(&refspec, 0, sizeof(refspec));
refspec.dst = (char *)refname;
- if (!remote_find_tracking(info->remote, &refspec)) {
- if (!((flags & REF_ISSYMREF) ||
- string_list_has_string(info->ref_names, refspec.src))) {
- struct ref *ref = make_linked_ref(refname, &info->stale_refs_tail);
- hashcpy(ref->new_sha1, sha1);
- }
+
+ /*
+ * If the user speicified refspecs on the command line, we
+ * should only use those to check. Otherwise, look in the
+ * remote's configuration for the branch.
+ */
+ if (info->ref_count)
+ ret = find_in_refs(info->refs, info->ref_count, &refspec);
+ else
+ ret = remote_find_tracking(info->remote, &refspec);
Minor typo in the comment. But more importantly, this feels like a very
low-level place to be thinking about things like what the user gave us
on the command line.
Shouldn't get_stale_heads not get a remote at all, and just get a set of
refspecs? Those should be the minimal information it needs to get its
answer, right?
OK, so take a step back and figure out what we want the rules to be
before we call get_stale_heads. It does sound like a more elegant
approach. I was trying to disrupt the callers as little as possible, but
then again, there's only two. Will change.
cmn
From: Jeff King <hidden> Date: 2016-06-15 22:52:12
On Fri, Oct 07, 2011 at 06:37:13PM +0200, Carlos Martín Nieto wrote:
quoted
I assume you mean s/tag/branch/ in the last line?
Yeah, maybe ref would be better?
Yeah, agreed.
quoted
Tests?
Good point.
It sounds like you already have a reproduction recipe for this, and for
the --tags thing in the next commit.
OK, so take a step back and figure out what we want the rules to be
before we call get_stale_heads. It does sound like a more elegant
approach. I was trying to disrupt the callers as little as possible, but
then again, there's only two. Will change.
Yeah. Sometimes we try hard to make a minimal patch, because it makes it
easier to review. At the same time, I think it's more important to make
the code clean if it needs it. Especially when there aren't many callers
to disrupt.
-Peff
@@ -699,8 +699,17 @@ static int do_fetch(struct transport *transport,free_refs(ref_map);return1;}-if(prune)+if(prune){+/* If --tags was specified, we need to tell prune_refs+*thatwe'refilteringtherefsfromtheremote*/+if(tags==TAGS_SET){+constchar*tags_refspec="refs/tags/*:refs/tags/*";+refs=xrealloc(refs,(ref_count+1)*sizeof(structrefspec));+refs[ref_count]=*parse_fetch_refspec(1,&tags_refspec);+ref_count++;+}prune_refs(transport,refs,ref_count,ref_map);+}
I don't think we can realloc refs here. It's passed into do_fetch. When
we realloc it, the old pointer value will be invalid. But when we return
from do_fetch, the caller (fetch_one) will still have that old value,
and will call free() on it.
Yes, you're right. I guess it's been working by luck and generous amount
of memory.
Instead, you have to make a whole new list, copy the old values in, add
your new one, and then free the result.