Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
Hi Jason,
Attached changes. I just edited the patch file directly.
Kind regards,
Anthony
From 7984020f16e715017e62b8637d2e69c1aec3478a Mon Sep 17 00:00:00 2001
From: Anthony Sharp
Date: Thu, 21 Jan 2021 15:26:25 +
Subject: [PATCH] c++: Private inheritance access diagnostics fix [PR17314]
This patch fixes PR17314. Previously, when class C attempted
to access member a declared in class A through class B, where class B
privately inherits from A and class C inherits from B, GCC would correctly
report an access violation, but would erroneously report that the reason was
because a was "protected", when in fact, from the point of view of class C,
it was really "private". This patch updates the diagnostics code to generate
more correct errors in cases of failed inheritance such as these.
The reason this bug happened was because GCC was examining the
declared access of decl, instead of looking at it in the
context of class inheritance.
gcc/cp/ChangeLog:
2021-01-21 Anthony Sharp
* call.c (complain_about_access): Altered function.
* cp-tree.h (complain_about_access): Changed parameters of function.
(get_parent_with_private_access): Declared new function.
* search.c (get_parent_with_private_access): Defined new function.
* semantics.c (enforce_access): Modified function.
* typeck.c (complain_about_unrecognized_member): Updated function
arguments in complain_about_access.
gcc/testsuite/ChangeLog:
2021-01-21 Anthony Sharp
* g++.dg/lookup/scoped1.C: Modified testcase to run successfully with
changes.
* g++.dg/tc1/dr142.C: Same as above.
* g++.dg/tc1/dr52.C: Same as above.
* g++.old-deja/g++.brendan/visibility6.C: Same as above.
* g++.old-deja/g++.brendan/visibility8.C: Same as above.
* g++.old-deja/g++.jason/access8.C: Same as above.
* g++.old-deja/g++.law/access4.C: Same as above.
* g++.old-deja/g++.law/visibility12.C: Same as above.
* g++.old-deja/g++.law/visibility4.C: Same as above.
* g++.old-deja/g++.law/visibility8.C: Same as above.
* g++.old-deja/g++.other/access4.C: Same as above.
---
gcc/cp/call.c | 38 ++-
gcc/cp/cp-tree.h | 4 +-
gcc/cp/search.c | 35 +
gcc/cp/semantics.c| 31 ++-
gcc/cp/typeck.c | 3 +-
gcc/testsuite/g++.dg/lookup/scoped1.C | 4 +-
gcc/testsuite/g++.dg/tc1/dr142.C | 8 ++--
gcc/testsuite/g++.dg/tc1/dr52.C | 6 +--
.../g++.old-deja/g++.brendan/visibility6.C| 4 +-
.../g++.old-deja/g++.brendan/visibility8.C| 4 +-
.../g++.old-deja/g++.jason/access8.C | 5 ++-
gcc/testsuite/g++.old-deja/g++.law/access4.C | 5 ++-
.../g++.old-deja/g++.law/visibility12.C | 7 ++--
.../g++.old-deja/g++.law/visibility4.C| 5 ++-
.../g++.old-deja/g++.law/visibility8.C| 5 ++-
.../g++.old-deja/g++.other/access4.C | 4 +-
16 files changed, 129 insertions(+), 39 deletions(-)
diff --git a/gcc/cp/call.c b/gcc/cp/call.c
index b6e9f125aeb..175a3d9b421 100644
--- a/gcc/cp/call.c
+++ b/gcc/cp/call.c
@@ -7142,27 +7142,45 @@ build_op_delete_call (enum tree_code code, tree addr, tree size,
/* Issue diagnostics about a disallowed access of DECL, using DIAG_DECL
in the diagnostics.
- If ISSUE_ERROR is true, then issue an error about the
- access, followed by a note showing the declaration.
- Otherwise, just show the note. */
+ If ISSUE_ERROR is true, then issue an error about the access, followed
+ by a note showing the declaration. Otherwise, just show the note.
+
+ DIAG_DECL and DIAG_LOCATION will almost always be the same.
+ DIAG_LOCATION is just another DECL. NO_ACCESS_REASON is an optional
+ parameter used to specify why DECL wasn't accessible (e.g. ak_private
+ would be because DECL was private). If not using NO_ACCESS_REASON,
+ then it must be ak_none, and the access failure reason will be
+ figured out by looking at the protection of DECL. */
void
-complain_about_access (tree decl, tree diag_decl, bool issue_error)
+complain_about_access (tree decl, tree diag_decl, tree diag_location,
+bool issue_error, access_kind no_access_reason)
{
- if (TREE_PRIVATE (decl))
+ /* If we have not already figured out why DECL is inaccessible... */
+ if (no_access_reason == ak_none)
+{
+ /* Examine the access of DECL to find out why. */
+ if (TREE_PRIVATE (decl))
+ no_access_reason = ak_private;
+ else if (TREE_PROTECTED (decl))
+ no_access_reason = ak_protected;
+}
+
+ /* Now generate an error message depending on calculated access. */
+ if (no_access_reason == ak_private)
{
if (issue_error)
error ("%q#D is private within this context", diag_decl);
- inform (DECL_SOURCE_LOCATION (diag_decl),
- "declared private here");
+ inform (DECL_SOURCE_LOCATION (diag_location),
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
On 1/22/21 3:07 PM, Anthony Sharp wrote:
Hi Jason,
Thanks for getting back to me so quickly.
> Why two gcc-comit-mklog? That would generate the log entries twice.
It did in fact generate the log entries twice, but I deleted out the
second copy. Perhaps it would have made more sense to do git commit
--amend instead.
> Instead of making access_in_type non-static, let's defiine
> get_parent_with_private_access in search.c and declare it in cp-tree.h
> (with the declarations of nearby search.c functions).
Done.
> Only one 'n' in inaccessible.
Oops!
Subsequent lines of a comment should be indented to line up with the
first line. This applies to all your multi-line comments.
My bad, hopefully fixed now.
Don't change the indentation of these blocks; in the GNU coding style
the { } are indented two spaces from the if.
I think I see what you mean; I forgot to indent the { } (and therefore
also everything within it, by two spaces). Hopefully fixed.
The new line of arguments should be indented to line up with the
first one.
Fixed I think.
Please find attached the latest patch version with all these changes.
git gcc-verify returns no problems and check_GNU_style.sh returns only
false positives. Source builds fine. To be super safe I re-cloned the
source and did git apply with the patch and it built and worked just
fine, and hopefully I haven't missed anything.
Thanks again for your help.
Subject: [PATCH] This patch fixes PR17314. Previously, when class C attempted
to access member a declared in class A through class B, where class B
privately inherits from A and class C inherits from B, GCC would correctly
report an access violation, but would erroneously report that the reason was
because a was "protected", when in fact, from the point of view of class C,
it was really "private". This patch updates the diagnostics code to generate
more correct errors in cases of failed inheritance such as these.
The first line of the commit message should be the subject line for the
patch, i.e. "c++: private inheritance access diagnostics fix [PR17314]",
then a blank line, then the rationale.
+ if (parent_binfo != NULL_TREE
+ && context_for_name_lookup (decl)
+ != BINFO_TYPE (parent_binfo))
Here you want parens around the second != expression and its != token
aligned with "context"
+ complain_about_access (decl, diag_decl, diag_location, true,
+ parent_access);
...
+ complain_about_access (afi.get_decl (), afi.get_diag_decl (),
+ afi.get_diag_decl (), false, ak_none);
In both these calls, the second line is indented one space too far.
Jason
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
Hi Jason,
Thanks for getting back to me so quickly.
> Why two gcc-comit-mklog? That would generate the log entries twice.
It did in fact generate the log entries twice, but I deleted out the second
copy. Perhaps it would have made more sense to do git commit --amend
instead.
> Instead of making access_in_type non-static, let's defiine
> get_parent_with_private_access in search.c and declare it in cp-tree.h
> (with the declarations of nearby search.c functions).
Done.
> Only one 'n' in inaccessible.
Oops!
Subsequent lines of a comment should be indented to line up with the
> first line. This applies to all your multi-line comments.
My bad, hopefully fixed now.
Don't change the indentation of these blocks; in the GNU coding style
> the { } are indented two spaces from the if.
>
I think I see what you mean; I forgot to indent the { } (and therefore also
everything within it, by two spaces). Hopefully fixed.
The new line of arguments should be indented to line up with the first one.
>
Fixed I think.
Please find attached the latest patch version with all these changes. git
gcc-verify returns no problems and check_GNU_style.sh returns only false
positives. Source builds fine. To be super safe I re-cloned the source and
did git apply with the patch and it built and worked just fine, and
hopefully I haven't missed anything.
Thanks again for your help.
Kind regards,
Anthony
From 7984020f16e715017e62b8637d2e69c1aec3478a Mon Sep 17 00:00:00 2001
From: Anthony Sharp
Date: Thu, 21 Jan 2021 15:26:25 +
Subject: [PATCH] This patch fixes PR17314. Previously, when class C attempted
to access member a declared in class A through class B, where class B
privately inherits from A and class C inherits from B, GCC would correctly
report an access violation, but would erroneously report that the reason was
because a was "protected", when in fact, from the point of view of class C,
it was really "private". This patch updates the diagnostics code to generate
more correct errors in cases of failed inheritance such as these.
The reason this bug happened was because GCC was examining the
declared access of decl, instead of looking at it in the
context of class inheritance.
gcc/cp/ChangeLog:
2021-01-21 Anthony Sharp
* call.c (complain_about_access): Altered function.
* cp-tree.h (complain_about_access): Changed parameters of function.
(get_parent_with_private_access): Declared new function.
* search.c (get_parent_with_private_access): Defined new function.
* semantics.c (enforce_access): Modified function.
* typeck.c (complain_about_unrecognized_member): Updated function
arguments in complain_about_access.
gcc/testsuite/ChangeLog:
2021-01-21 Anthony Sharp
* g++.dg/lookup/scoped1.C: Modified testcase to run successfully with
changes.
* g++.dg/tc1/dr142.C: Same as above.
* g++.dg/tc1/dr52.C: Same as above.
* g++.old-deja/g++.brendan/visibility6.C: Same as above.
* g++.old-deja/g++.brendan/visibility8.C: Same as above.
* g++.old-deja/g++.jason/access8.C: Same as above.
* g++.old-deja/g++.law/access4.C: Same as above.
* g++.old-deja/g++.law/visibility12.C: Same as above.
* g++.old-deja/g++.law/visibility4.C: Same as above.
* g++.old-deja/g++.law/visibility8.C: Same as above.
* g++.old-deja/g++.other/access4.C: Same as above.
---
gcc/cp/call.c | 38 ++-
gcc/cp/cp-tree.h | 4 +-
gcc/cp/search.c | 35 +
gcc/cp/semantics.c| 31 ++-
gcc/cp/typeck.c | 3 +-
gcc/testsuite/g++.dg/lookup/scoped1.C | 4 +-
gcc/testsuite/g++.dg/tc1/dr142.C | 8 ++--
gcc/testsuite/g++.dg/tc1/dr52.C | 6 +--
.../g++.old-deja/g++.brendan/visibility6.C| 4 +-
.../g++.old-deja/g++.brendan/visibility8.C| 4 +-
.../g++.old-deja/g++.jason/access8.C | 5 ++-
gcc/testsuite/g++.old-deja/g++.law/access4.C | 5 ++-
.../g++.old-deja/g++.law/visibility12.C | 7 ++--
.../g++.old-deja/g++.law/visibility4.C| 5 ++-
.../g++.old-deja/g++.law/visibility8.C| 5 ++-
.../g++.old-deja/g++.other/access4.C | 4 +-
16 files changed, 129 insertions(+), 39 deletions(-)
diff --git a/gcc/cp/call.c b/gcc/cp/call.c
index b6e9f125aeb..175a3d9b421 100644
--- a/gcc/cp/call.c
+++ b/gcc/cp/call.c
@@ -7142,27 +7142,45 @@ build_op_delete_call (enum tree_code code, tree addr, tree size,
/* Issue diagnostics about a disallowed access of DECL, using DIAG_DECL
in the diagnostics.
- If ISSUE_ERROR is true, then issue an error about the
- access, followed by a note showing the declaration.
- Otherwise, just show the note. */
+ If ISSUE_ERROR is true, then issue an error about the access, followed
+ by a note showing the declaration. Otherwise, just show the note.
+
+ DIAG_DECL and DIAG_LOCATION will almost always be the same.
+ DIAG_LOCATION
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
On 1/21/21 2:28 PM, Anthony Sharp wrote:
Hi Jason,
I've finally completed my copyright assignment form. I've attached it
to this email for reference.
You don't need write access to the main repository to use these commands
on your local copy. One nice thing about git compared to svn is that
you don't need to touch the server for anything but push and pull.
Incidentally, how are you producing your patch? Maybe try git
format-patch instead.
The method I am using at the moment is the one Ranjit Mathew talks
about here: http://rmathew.com/articles/gcj/crpatch.html. Actually,
having just re-read it, it says: 'NOTE: This is not the “proper” or
“official” way of creating and submitting patches - that process has
been explained in detail elsewhere. That process requires one to use
Subversion (SVN). The process described here is meant for “one-off
hackers” or people who cannot use SVN for some reason or the other.'
... oops!
It's my fault kind of - the official GCC webpage
(https://gcc.gnu.org/gitwrite.html) explaining how to do it is called
'Read-write Git access' so I assumed it was only relevant for people
who have access to the repo, but I see that is not the case.
I've tried the git way of doing it and I'm attaching a new patch file
that (hopefully) is better this time. Basically what I did was what
you suggested:
git pull
contrib/gcc-git-customization.sh
(make changes)
git add *
git gcc-commit-mklog
git gcc-commit-mklog --amend
git format-patch -1 master
I also re-built the source just to make sure I hadn't messed anything
up. I re-ran the C++ regression tests using make check-c and make
check-c++. Whilst I did not do a before/after comparison of the
results, I checked the FAILs in gcc.sum and g++.sum and they all
looked like they had nothing to do with my code. All the code is the
same as before, so I'm thinking it should be fine (I just wanted to be
safe). Also checked against check_GNU_style.sh.
Assuming that's all fine, as for the code itself, there might well be
some tweaks that could make it better, and so if that is the case then
please let me know.
The code looks good, I just have some minor tweaks. Thanks!
+++ b/gcc/cp/semantics.c
...
+extern access_kind access_in_type (tree type, tree decl);
...
+static tree
+get_parent_with_private_access (tree decl, tree binfo)
Instead of making access_in_type non-static, let's defiine
get_parent_with_private_access in search.c and declare it in cp-tree.h
(with the declarations of nearby search.c functions).
+ /* If we have not already figured out why DECL is innaccessible... */
...
+ /* Couldn't figure out why DECL is innaccesible, so just say it's
+ innaccessible. */
Only one 'n' in inaccessible.
There are various minor formatting issues:
(https://www.gnu.org/prep/standards/standards.html#Formatting)
+ /* Couldn't figure out why DECL is innaccesible, so just say it's
+ innaccessible. */
Subsequent lines of a comment should be indented to line up with the
first line. This applies to all your multi-line comments.
-{
- if (issue_error)
- error ("%q#D is private within this context", diag_decl);
- inform (DECL_SOURCE_LOCATION (diag_decl),
- "declared private here");
-}
+ {
+if (issue_error)
+ error ("%q#D is private within this context", diag_decl);
+inform (DECL_SOURCE_LOCATION (diag_location), "declared private here");
+ }
Don't change the indentation of these blocks; in the GNU coding style
the { } are indented two spaces from the if.
+ tree parent_binfo = get_parent_with_private_access (decl,
+ basetype_path);
...
+ complain_about_access (decl, diag_decl, diag_location, true,
+ parent_access);
The new line of arguments should be indented to line up with the first one.
Jason
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
On 1/21/21 2:28 PM, Anthony Sharp wrote: Hi Jason, I've finally completed my copyright assignment form. I've attached it to this email for reference. You don't need write access to the main repository to use these commands on your local copy. One nice thing about git compared to svn is that you don't need to touch the server for anything but push and pull. Incidentally, how are you producing your patch? Maybe try git format-patch instead. The method I am using at the moment is the one Ranjit Mathew talks about here: http://rmathew.com/articles/gcj/crpatch.html. Interesting. Yeah, that page is obsolete since the move to git. It's my fault kind of - the official GCC webpage (https://gcc.gnu.org/gitwrite.html) explaining how to do it is called 'Read-write Git access' so I assumed it was only relevant for people who have access to the repo, but I see that is not the case. Those pages could definitely be more clearly organized. I've tried the git way of doing it and I'm attaching a new patch file that (hopefully) is better this time. Basically what I did was what you suggested: git pull contrib/gcc-git-customization.sh (make changes) git add * git gcc-commit-mklog git gcc-commit-mklog --amend Why two gcc-comit-mklog? That would generate the log entries twice. You should also git gcc-verify at this point; for me, it complains about some of your header lines in the log. Your author line needs to start at the first column, and use "01" for January instead of just "1". The other explanatory lines can be omitted, in favor of: The commit message before the log entries should include your rationale for the patch (e.g. the first two paragraphs of your initial email). git format-patch -1 master I also re-built the source just to make sure I hadn't messed anything up. I re-ran the C++ regression tests using make check-c and make check-c++. Whilst I did not do a before/after comparison of the results, I checked the FAILs in gcc.sum and g++.sum and they all looked like they had nothing to do with my code. All the code is the same as before, so I'm thinking it should be fine (I just wanted to be safe). Also checked against check_GNU_style.sh. Assuming that's all fine, as for the code itself, there might well be some tweaks that could make it better, and so if that is the case then please let me know. I'll look at the code soon. Thanks, Jason
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
Hi Jason, I've finally completed my copyright assignment form. I've attached it to this email for reference. > You don't need write access to the main repository to use these commands > on your local copy. One nice thing about git compared to svn is that > you don't need to touch the server for anything but push and pull. > > Incidentally, how are you producing your patch? Maybe try git > format-patch instead. The method I am using at the moment is the one Ranjit Mathew talks about here: http://rmathew.com/articles/gcj/crpatch.html. Actually, having just re-read it, it says: 'NOTE: This is not the “proper” or “official” way of creating and submitting patches - that process has been explained in detail elsewhere. That process requires one to use Subversion (SVN). The process described here is meant for “one-off hackers” or people who cannot use SVN for some reason or the other.' ... oops! It's my fault kind of - the official GCC webpage (https://gcc.gnu.org/gitwrite.html) explaining how to do it is called 'Read-write Git access' so I assumed it was only relevant for people who have access to the repo, but I see that is not the case. I've tried the git way of doing it and I'm attaching a new patch file that (hopefully) is better this time. Basically what I did was what you suggested: git pull contrib/gcc-git-customization.sh (make changes) git add * git gcc-commit-mklog git gcc-commit-mklog --amend git format-patch -1 master I also re-built the source just to make sure I hadn't messed anything up. I re-ran the C++ regression tests using make check-c and make check-c++. Whilst I did not do a before/after comparison of the results, I checked the FAILs in gcc.sum and g++.sum and they all looked like they had nothing to do with my code. All the code is the same as before, so I'm thinking it should be fine (I just wanted to be safe). Also checked against check_GNU_style.sh. Assuming that's all fine, as for the code itself, there might well be some tweaks that could make it better, and so if that is the case then please let me know. Kind regards, Anthony Sharp Sharp.1672871.GCC.pdf Description: Adobe PDF document From 9f7fa0b4a892f717974d79a6a56a5f8a8a8d9943 Mon Sep 17 00:00:00 2001 From: Anthony Sharp Date: Thu, 21 Jan 2021 15:26:25 + Subject: [PATCH] gcc/cp/ChangeLog: 2021-1-21 Anthony Sharp Fixes PR17314 * call.c (complain_about_access): Altered function. * cp-tree.h (complain_about_access): Changed parameters of function. * search.c (access_in_type): Made function non-static so it can be used in semantics.c. * semantics.c (access_in_type): Added as extern function. (get_parent_with_private_access): Added function. (enforce_access): Modified function. * typeck.c (complain_about_unrecognized_member): Updated function arguments in complain_about_access. gcc/testsuite/ChangeLog: 2021-1-21 Anthony Sharp Changes required due to PR17314 fix * g++.dg/lookup/scoped1.C: Modified testcase to run successfully with changes. * g++.dg/tc1/dr142.C: Same as above. * g++.dg/tc1/dr52.C: Same as above. * g++.old-deja/g++.brendan/visibility6.C: Same as above. * g++.old-deja/g++.brendan/visibility8.C: Same as above. * g++.old-deja/g++.jason/access8.C: Same as above. * g++.old-deja/g++.law/access4.C: Same as above. * g++.old-deja/g++.law/visibility12.C: Same as above. * g++.old-deja/g++.law/visibility4.C: Same as above. * g++.old-deja/g++.law/visibility8.C: Same as above. * g++.old-deja/g++.other/access4.C: Same as above. --- gcc/cp/call.c | 64 ++--- gcc/cp/cp-tree.h | 3 +- gcc/cp/search.c | 4 +- gcc/cp/semantics.c| 68 ++- gcc/cp/typeck.c | 3 +- gcc/testsuite/g++.dg/lookup/scoped1.C | 4 +- gcc/testsuite/g++.dg/tc1/dr142.C | 8 +-- gcc/testsuite/g++.dg/tc1/dr52.C | 6 +- .../g++.old-deja/g++.brendan/visibility6.C| 4 +- .../g++.old-deja/g++.brendan/visibility8.C| 4 +- .../g++.old-deja/g++.jason/access8.C | 5 +- gcc/testsuite/g++.old-deja/g++.law/access4.C | 5 +- .../g++.old-deja/g++.law/visibility12.C | 7 +- .../g++.old-deja/g++.law/visibility4.C| 5 +- .../g++.old-deja/g++.law/visibility8.C| 5 +- .../g++.old-deja/g++.other/access4.C | 4 +- 16 files changed, 145 insertions(+), 54 deletions(-) diff --git a/gcc/cp/call.c b/gcc/cp/call.c index b6e9f125aeb..fc2f1c6226c 100644 --- a/gcc/cp/call.c +++ b/gcc/cp/call.c @@ -7142,33 +7142,51 @@ build_op_delete_call (enum tree_code code, tree addr, tree size, /* Issue diagnostics about a disallowed access of DECL, using DIAG_DECL in the diagnostics. - If ISSUE_ERROR is true, then issue an error about the - access, followed by a note showing the declaration. - Otherwise, just show the note. */ + If ISSUE_ERROR is true, then issue an error about
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
On 1/8/21 7:38 PM, Anthony Sharp wrote: Hi Jason, Thank you! To start with, do you have a copyright assignment on file or in the works already? Good point. I incorrectly assumed it would only be a minor contribution copyright-wise. > Mr Edelsohn gave me a template which I've now filled out and sent to ass...@gnu.org. I'm assuming I just need to wait for them to send me the form. I'll update this thread when that's sorted. In the meantime I've hopefully fixed some of the issues. Great. Second, your patch was mangled by word wrap so that it can't be applied without manual repair. If you can't prevent word wrap in your mail client, please send it as an attachment rather than inline. Oh yes I see where it's gotten mangled now. I'm attaching it as a .patch file (I assume that's okay). Also, there are a few whitespace issues in the patch; please run contrib/check_GNU_style.sh on the patch before submitting. Should be all fixed now (there is one style issue left but it's a false positive). Visual Studio Code was lying to me about what the file looks like so if there are any more formatting issues please let me know. If you use contrib/gcc-git-customization.sh and then git gcc-commit-mklog you don't need to touch ChangeLog files at all, just adjust the generated ChangeLog entries in the git commit message. I personally tend to commit first with a placeholder message and then use git gcc-commit-mklog --amend to generate the ChangeLog entries. Wouldn't that require read-write access? (Just from looking here https://gcc.gnu.org/gitwrite.html.) You don't need write access to the main repository to use these commands on your local copy. One nice thing about git compared to svn is that you don't need to touch the server for anything but push and pull. Incidentally, how are you producing your patch? Maybe try git format-patch instead. Probably. Can you use sort/uniq/diff on the .sum testsuite output to determine which passes are missing in the patched sources? According to contrib/dg-cmp-results.sh ... I get a bunch of these weird NA->PASSes (and vice-versa), for example: PASS->NA: g++.dg/modules/alias-1_a.H module-cmi (gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_clean/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm) NA->PASS: g++.dg/modules/alias-1_a.H module-cmi (gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_pr17314/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm) PASS->NA: g++.dg/modules/alias-1_a.H module-cmi (gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_clean/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm) NA->PASS: g++.dg/modules/alias-1_a.H module-cmi (gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_pr17314/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm) They're weird because I haven't actually touched those files (so I'm assuming this is normal). There are about ~400 of those and they're all .gcm files. They seem to balance out. The modules code and tests are very new and volatile, I wouldn't worry about them. dr142.c reports: NA->PASS: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 11) PASS->NA: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 5) PASS->NA: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 7) PASS->NA: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 8) NA->PASS: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 11) PASS->NA: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 5) PASS->NA: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 7) PASS->NA: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 8) NA->PASS: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 11) PASS->NA: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 5) PASS->NA: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 7) PASS->NA: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 8) NA->PASS: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 11) PASS->NA: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 5) PASS->NA: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 7) PASS->NA: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 8) These changes are because your patch changes that test to expect warnings in different places. In other words, there are 12 PASS->NAs and 4 NA->PASSes in this file, meaning a net change of -8 (which explains why there are eight fewer). My other changes also report PASS->NAs and vice-versa, but for those the number of new NAs equals the number of new PASSes, so they don't cause a change in quantity. Thanks for being patient with me. I'll let you know when I've completed the forms. Also if I need to adjust the .patch to deal with the changelogs issue please let me know. Kind regards, Anthony
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
Hi Jason,
Thank you!
> To start with, do you have a copyright assignment on file or in the
> works already?
Good point. I incorrectly assumed it would only be a minor
contribution copyright-wise. Mr Edelsohn gave me a template which I've
now filled out and sent to ass...@gnu.org. I'm assuming I just need to
wait for them to send me the form. I'll update this thread when that's
sorted. In the meantime I've hopefully fixed some of the issues.
> Second, your patch was mangled by word wrap so that it can't be applied
> without manual repair. If you can't prevent word wrap in your mail
> client, please send it as an attachment rather than inline.
Oh yes I see where it's gotten mangled now. I'm attaching it as a
.patch file (I assume that's okay).
> Also, there are a few whitespace issues in the patch; please run
> contrib/check_GNU_style.sh on the patch before submitting.
Should be all fixed now (there is one style issue left but it's a
false positive). Visual Studio Code was lying to me about what the
file looks like so if there are any more formatting issues please let
me know.
> If you use contrib/gcc-git-customization.sh and then git
> gcc-commit-mklog you don't need to touch ChangeLog files at all, just
> adjust the generated ChangeLog entries in the git commit message. I
> personally tend to commit first with a placeholder message and then use
> git gcc-commit-mklog --amend to generate the ChangeLog entries.
Wouldn't that require read-write access? (Just from looking here
https://gcc.gnu.org/gitwrite.html.)
> Probably. Can you use sort/uniq/diff on the .sum testsuite output to
> determine which passes are missing in the patched sources?
According to contrib/dg-cmp-results.sh ...
I get a bunch of these weird NA->PASSes (and vice-versa), for example:
PASS->NA: g++.dg/modules/alias-1_a.H module-cmi
(gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_clean/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm)
NA->PASS: g++.dg/modules/alias-1_a.H module-cmi
(gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_pr17314/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm)
PASS->NA: g++.dg/modules/alias-1_a.H module-cmi
(gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_clean/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm)
NA->PASS: g++.dg/modules/alias-1_a.H module-cmi
(gcm.cache/home/anthony/Desktop/GCC/builds_and_source/source_pr17314/gcc/testsuite/g++.dg/modules/alias-1_a.H.gcm)
They're weird because I haven't actually touched those files (so I'm
assuming this is normal). There are about ~400 of those and they're
all .gcm files. They seem to balance out.
dr142.c reports:
NA->PASS: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 11)
PASS->NA: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 5)
PASS->NA: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 7)
PASS->NA: g++.dg/tc1/dr142.C -std=c++14 (test for warnings, line 8)
NA->PASS: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 11)
PASS->NA: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 5)
PASS->NA: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 7)
PASS->NA: g++.dg/tc1/dr142.C -std=c++17 (test for warnings, line 8)
NA->PASS: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 11)
PASS->NA: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 5)
PASS->NA: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 7)
PASS->NA: g++.dg/tc1/dr142.C -std=c++2a (test for warnings, line 8)
NA->PASS: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 11)
PASS->NA: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 5)
PASS->NA: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 7)
PASS->NA: g++.dg/tc1/dr142.C -std=c++98 (test for warnings, line 8)
In other words, there are 12 PASS->NAs and 4 NA->PASSes in this file,
meaning a net change of -8 (which explains why there are eight fewer).
My other changes also report PASS->NAs and vice-versa, but for those
the number of new NAs equals the number of new PASSes, so they don't
cause a change in quantity.
Thanks for being patient with me. I'll let you know when I've
completed the forms.
Also if I need to adjust the .patch to deal with the changelogs issue
please let me know.
Kind regards,
Anthony
Index: gcc/testsuite/g++.old-deja/g++.jason/access8.C
===
--- gcc/testsuite/g++.old-deja/g++.jason/access8.C 2020-12-31 16:51:34.0 +
+++ gcc/testsuite/g++.old-deja/g++.jason/access8.C 2021-01-03 00:22:14.969854000 +
@@ -4,5 +4,5 @@
// Bug: g++ forgets access decls after the definition.
-class inh { // { dg-message "" } inaccessible
+class inh {
int a;
protected:
@@ -10,5 +10,6 @@ protected:
};
-class mel : private inh {
+class mel : private inh // { dg-message "" } inaccessible
+{
protected:
int t;
Index: gcc/testsuite/g++.old-deja/g++.law/access4.C
Re: [PATCH] c++: private inheritance access diagnostics fix [PR17314]
On 1/5/21 9:24 AM, Anthony Sharp via Gcc-patches wrote: This patch fixes PR17314 (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17314). Previously, when class C attempted to access member a declared in class A through class B, where class B privately inherits from A and class C inherits from B, GCC would correctly report an access violation, but would erroneously report that the reason was because a was "protected", when in fact, from the point of view of class C, it was "private". This patch updates the diagnostics code to generate more correct errors in cases of failed inheritance such as these. The reason this bug happened was because GCC was examining the declared access of decl, instead of looking at it in the context of class inheritance. --- COMMENTS --- This is my first GCC patch ever Thanks, and welcome! To start with, do you have a copyright assignment on file or in the works already? If not, see the top of https://gcc.gnu.org/contribute.html for more information. I probably shouldn't look at the patch in detail until that's addressed. Second, your patch was mangled by word wrap so that it can't be applied without manual repair. If you can't prevent word wrap in your mail client, please send it as an attachment rather than inline. Also, there are a few whitespace issues in the patch; please run contrib/check_GNU_style.sh on the patch before submitting. so there is probably something I have done very wrong. Please let me know :) The thought of my code being scrutinised by people with PhDs and doctorates is quite frankly terrifying. Note that since it is a new year I had to make a new changelog file so the diff for the patch might be slightly off. If you use contrib/gcc-git-customization.sh and then git gcc-commit-mklog you don't need to touch ChangeLog files at all, just adjust the generated ChangeLog entries in the git commit message. I personally tend to commit first with a placeholder message and then use git gcc-commit-mklog --amend to generate the ChangeLog entries. There was no need to add additional regression tests since it was adequate to simply change some of the regression tests that were there originally (all the patch changes is the informative message telling the user where a decl was defined as private). Agreed. --- REGRESSION ANALYSIS --- No regressions reported. G++ (CLEAN) RESULTS # of expected passes202879 # of unexpected failures1 # of expected failures988 # of unsupported tests8654 GCC (CLEAN) RESULTS # of expected passes163377 # of unexpected failures94 # of unexpected successes37 # of expected failures915 # of unsupported tests2530 G++ (PR17314 PATCHED) RESULTS # of expected passes202871 # of unexpected failures1 # of expected failures988 # of unsupported tests8654 GCC (PR17314 PATCHED) RESULTS # of expected passes163377 # of unexpected failures94 # of unexpected successes37 # of expected failures915 # of unsupported tests2530 When I build and make -k check -j 6 on the patched source it reports 202871 passes (8 fewer), although the FAILs do not increase. I am not 100% sure why this happens since I have not removed any testcases, only edited a few, but I think this happens because in files like dr142.c I removed more output checks than I added. make -k check -j 6 also returns error 2 sometimes, although there are no obvious errors or warnings in the logs explaining why. Probably harmless? Probably. Can you use sort/uniq/diff on the .sum testsuite output to determine which passes are missing in the patched sources? --- BUILD REPORT --- GCC builds normally on x86_64-pc-linux-gnu for x86_64-pc-linux-gnu using make -j 6. I didn't see it necessary to test on other build targets since the patch only affects the C++ front end and so functionality is unlikely to differ between platforms. The compile log reports: Comparing stages 2 and 3 warning: gcc/cc1obj-checksum.o differs Comparison successful. and then continues. I assume this means it was actually successful. Yes. Thanks, Jason Index: gcc/cp/ChangeLog from Anthony Sharp Fixes PR17314 * typeck.c (complain_about_unrecognized_member): Updated function arguments in complain_about_access. * call.c (complain_about_access): Altered function. * semantics.c (get_parent_with_private_access): Added function. (access_in_type): Added as extern function. * search.c (access_in_type): Made function non-static so it can be used in semantics.c. * cp-tree.h (complain_about_access): Changed parameters of function. Index: gcc/testsuite/ChangeLog from Anthony Sharp Fixes PR17314 * g++.dg/lookup/scoped1.c modified testcase to run successfully with changes. * g++.dg/tc1/dr142.c modified testcase to run successfully with changes. *
[PATCH] c++: private inheritance access diagnostics fix [PR17314]
This patch fixes PR17314 (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17314). Previously, when class C attempted to access member a declared in class A through class B, where class B privately inherits from A and class C inherits from B, GCC would correctly report an access violation, but would erroneously report that the reason was because a was "protected", when in fact, from the point of view of class C, it was "private". This patch updates the diagnostics code to generate more correct errors in cases of failed inheritance such as these. The reason this bug happened was because GCC was examining the declared access of decl, instead of looking at it in the context of class inheritance. --- COMMENTS --- This is my first GCC patch ever so there is probably something I have done very wrong. Please let me know :) The thought of my code being scrutinised by people with PhDs and doctorates is quite frankly terrifying. Note that since it is a new year I had to make a new changelog file so the diff for the patch might be slightly off. There was no need to add additional regression tests since it was adequate to simply change some of the regression tests that were there originally (all the patch changes is the informative message telling the user where a decl was defined as private). --- REGRESSION ANALYSIS --- No regressions reported. G++ (CLEAN) RESULTS # of expected passes202879 # of unexpected failures1 # of expected failures988 # of unsupported tests8654 GCC (CLEAN) RESULTS # of expected passes163377 # of unexpected failures94 # of unexpected successes37 # of expected failures915 # of unsupported tests2530 G++ (PR17314 PATCHED) RESULTS # of expected passes202871 # of unexpected failures1 # of expected failures988 # of unsupported tests8654 GCC (PR17314 PATCHED) RESULTS # of expected passes163377 # of unexpected failures94 # of unexpected successes37 # of expected failures915 # of unsupported tests2530 When I build and make -k check -j 6 on the patched source it reports 202871 passes (8 fewer), although the FAILs do not increase. I am not 100% sure why this happens since I have not removed any testcases, only edited a few, but I think this happens because in files like dr142.c I removed more output checks than I added. make -k check -j 6 also returns error 2 sometimes, although there are no obvious errors or warnings in the logs explaining why. Probably harmless? --- BUILD REPORT --- GCC builds normally on x86_64-pc-linux-gnu for x86_64-pc-linux-gnu using make -j 6. I didn't see it necessary to test on other build targets since the patch only affects the C++ front end and so functionality is unlikely to differ between platforms. The compile log reports: Comparing stages 2 and 3 warning: gcc/cc1obj-checksum.o differs Comparison successful. and then continues. I assume this means it was actually successful. Index: gcc/cp/ChangeLog from Anthony Sharp Fixes PR17314 * typeck.c (complain_about_unrecognized_member): Updated function arguments in complain_about_access. * call.c (complain_about_access): Altered function. * semantics.c (get_parent_with_private_access): Added function. (access_in_type): Added as extern function. * search.c (access_in_type): Made function non-static so it can be used in semantics.c. * cp-tree.h (complain_about_access): Changed parameters of function. Index: gcc/testsuite/ChangeLog from Anthony Sharp Fixes PR17314 * g++.dg/lookup/scoped1.c modified testcase to run successfully with changes. * g++.dg/tc1/dr142.c modified testcase to run successfully with changes. * g++.dg/tc1/dr142.c modified testcase to run successfully with changes. * g++.dg/tc1/dr142.c modified testcase to run successfully with changes. * g++.dg/tc1/dr52.c modified testcase to run successfully with changes. * g++.old-deja/g++.brendan/visibility6.c modified testcase to run successfully with changes. * g++.old-deja/g++.brendan/visibility8.c modified testcase to run successfully with changes. * g++.old-deja/g++.jason/access8.c modified testcase to run successfully with changes. * g++.old-deja/g++.law/access4.c modified testcase to run successfully with changes. * g++.old-deja/g++.law/visibility12.c modified testcase to run successfully with changes. * g++.old-deja/g++.law/visibility4.c modified testcase to run successfully with changes. * g++.old-deja/g++.law/visibility8.c modified testcase to run successfully with changes. * g++.old-deja/g++.other/access4.c modified testcase to run successfully with changes. Index: gcc/testsuite/g++.old-deja/g++.jason/access8.C === --- gcc/testsuite/g++.old-deja/g++.jason/access8.C
